lang: core: embedded: Add standalone provisioning tool

This commit adds the ability to build a standalone provisioning tool.
This is the first useful public mcl code base as well. It is not
perfect, but does serve as a rough starting point to show what is
possible. In the future as the language and the engine evolve, this will
likely get more elegant, and also grow new features.

To build this, run `make clean && GOTAGS='embedded_provisioner' make`.
To run this, run `mgmt provisioner`.

.ackrc

@@ -1,3 +1,6 @@
 --ignore-dir=old/
 --ignore-dir=tmp/
 --ignore-dir=vendor/
+--ignore-dir=releases/
+--ignore-dir=rpmbuild/
+--type-set=mcl:ext:mcl

.editorconfig

@@ -23,3 +23,6 @@ indent_style = tab
 
 [*.mcl]
 indent_style = tab
+
+[*.txtar]
+indent_style = tab

.github/FUNDING.yml

@@ -0,0 +1,5 @@
+# You can add one username per supported platform and one custom link.
+custom: "https://paypal.me/purpleidea"
+github: purpleidea
+liberapay: purpleidea
+patreon: purpleidea

.github/settings.yml

@@ -68,6 +68,8 @@ labels:
     color: e11d21
   - name: question
     color: cc317c
+  - name: needinfo
+    color: fbca04
   - name: wontfix
     color: ffffff
 #  - name: first-timers-only

.github/workflows/test.yaml

@@ -0,0 +1,68 @@
+# Docs: https://help.github.com/en/articles/workflow-syntax-for-github-actions
+
+# If the name is omitted, it uses the filename instead.
+#name: Test
+on:
+  # Run on all pull requests.
+  pull_request:
+    #branches:
+    #- master
+  # Run on all pushes.
+  push:
+  # Run daily at 4am.
+  schedule:
+  - cron: 0 4 * * *
+
+jobs:
+  maketest:
+    name: Test (${{ matrix.test_block }}) on ${{ matrix.os }} with golang ${{ matrix.golang_version }}
+    runs-on: ${{ matrix.os }}
+    env:
+      GOPATH: /home/runner/work/mgmt/mgmt/go
+    strategy:
+      matrix:
+        # TODO: Add tip when it's supported: https://github.com/actions/setup-go/issues/21
+        os:
+          - ubuntu-latest
+          # macos tests are currently failing in CI
+          #- macos-latest
+        golang_version:
+          # TODO: add 1.21.x and tip
+          # minimum required and latest published go_version
+          - "1.20"
+        test_block:
+          - basic
+          - shell
+          - race
+      fail-fast: false
+
+    steps:
+      # Do not shallow fetch. The path can't be absolute, so we need to move it
+      # to the expected location later.
+      - name: Clone mgmt
+        uses: actions/checkout@v4
+        with:
+          submodules: recursive
+          fetch-depth: 0
+          path: ./go/src/github.com/purpleidea/mgmt
+
+      - name: Install Go ${{ matrix.golang_version }}
+        uses: actions/setup-go@v4
+        with:
+          go-version: ${{ matrix.golang_version }}
+
+      # Install & configure ruby, fixes gem permissions error
+      - name: Install Ruby
+        uses: ruby/setup-ruby@v1
+        with:
+          ruby-version: head
+
+      - name: Install dependencies
+        working-directory: ./go/src/github.com/purpleidea/mgmt
+        run: |
+          make deps
+
+      - name: Run test
+        working-directory: ./go/src/github.com/purpleidea/mgmt
+        run: |
+          TEST_BLOCK="${{ matrix.test_block }}" make test

.gitignore

@@ -5,8 +5,8 @@
 .envrc
 old/
 tmp/
+*WIP
 *_stringer.go
-bindata/*.go
 mgmt
 mgmt.static
 # crossbuild artifacts
@@ -14,3 +14,7 @@ build/mgmt-*
 mgmt.iml
 rpmbuild/
 releases/
+# vim swap files
+.*.sw[op]
+# prevent `echo foo 2>1` typo errors by making this file read-only
+1

.gitmodules

@@ -1,33 +0,0 @@
-[submodule "vendor/github.com/coreos/etcd"]
-	path = vendor/github.com/coreos/etcd
-	url = https://github.com/coreos/etcd/
-[submodule "vendor/google.golang.org/grpc"]
-	path = vendor/google.golang.org/grpc
-	url = https://github.com/grpc/grpc-go
-[submodule "vendor/github.com/grpc-ecosystem/grpc-gateway"]
-	path = vendor/github.com/grpc-ecosystem/grpc-gateway
-	url = https://github.com/grpc-ecosystem/grpc-gateway
-[submodule "vendor/gopkg.in/fsnotify.v1"]
-	path = vendor/gopkg.in/fsnotify.v1
-	url = https://gopkg.in/fsnotify.v1
-[submodule "vendor/github.com/purpleidea/go-systemd"]
-	path = vendor/github.com/purpleidea/go-systemd
-	url = https://github.com/purpleidea/go-systemd
-[submodule "vendor/honnef.co/go/augeas"]
-	path = vendor/honnef.co/go/augeas
-	url = https://github.com/dominikh/go-augeas/
-[submodule "vendor/github.com/grpc-ecosystem/go-grpc-prometheus"]
-	path = vendor/github.com/grpc-ecosystem/go-grpc-prometheus
-	url = https://github.com/grpc-ecosystem/go-grpc-prometheus
-[submodule "vendor/github.com/ugorji/go"]
-	path = vendor/github.com/ugorji/go
-	url = https://github.com/ugorji/go
-[submodule "vendor/github.com/purpleidea/docker"]
-	path = vendor/github.com/docker/docker
-	url = https://github.com/purpleidea/docker
-[submodule "vendor/github.com/purpleidea/distribution"]
-	path = vendor/github.com/docker/distribution
-	url = https://github.com/purpleidea/distribution
-[submodule "vendor/github.com/purpleidea/go-connections"]
-	path = vendor/github.com/docker/go-connections
-	url = https://github.com/docker/go-connections

.travis.yml

@@ -1,17 +1,18 @@
 language: go
 os:
   - linux
-go:
-  - 1.9.x
-  - 1.10.x
-  - tip
 go_import_path: github.com/purpleidea/mgmt
 sudo: true
-dist: trusty
+dist: xenial
 # travis requires that you update manually, and provides this key to trigger it
 apt:
   update: true
 before_install:
+  # print some debug information to help catch the constant travis regressions
+  - if [ -e /etc/apt/sources.list.d/ ]; then sudo ls -l /etc/apt/sources.list.d/; fi
+  # workaround broken travis NO_PUBKEY errors
+  - if [ -e /etc/apt/sources.list.d/rabbitmq_rabbitmq-server.list ]; then sudo rm -f /etc/apt/sources.list.d/rabbitmq_rabbitmq-server.list; fi
+  - if [ -e /etc/apt/sources.list.d/github_git-lfs.list ]; then sudo rm -f /etc/apt/sources.list.d/github_git-lfs.list; fi
   # as per a number of comments online, this might mitigate some flaky fails...
   - if [[ "$TRAVIS_OS_NAME" != "osx" ]]; then sudo apt-key adv --keyserver hkp://keyserver.ubuntu.com:80 --recv 0C49F3730359A14518585931BC711F9BA15703C6; fi
   # apt update tends to be flaky in travis, retry up to 3 times on failure
@@ -20,24 +21,34 @@ before_install:
   - git config remote.origin.fetch "+refs/heads/*:refs/remotes/origin/*"
   - git fetch --unshallow
 install: 'make deps'
-script: 'make test'
 matrix:
   fast_finish: false
   allow_failures:
-    - go: 1.10.x
-    - go: tip
-    - os: osx
+  - go: 1.21.x
+  - go: tip
+  - os: osx
   # include only one build for osx for a quicker build as the nr. of these runners are sparse
   include:
-    - os: osx
-      go: 1.9.x
+  - name: "basic tests"
+    go: 1.20.x
+    env: TEST_BLOCK=basic
+  - name: "shell tests"
+    go: 1.20.x
+    env: TEST_BLOCK=shell
+  - name: "race tests"
+    go: 1.20.x
+    env: TEST_BLOCK=race
+  - go: 1.21.x
+  - go: tip
+  - os: osx
+script: 'TEST_BLOCK="$TEST_BLOCK" make test'
 
 # the "secure" channel value is the result of running: ./misc/travis-encrypt.sh
 # with a value of: irc.freenode.net#mgmtconfig to eliminate noise from forks...
 notifications:
   irc:
-    channels:
-      - secure: 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
+    #channels:
+    #  - secure: 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
     template:
       - "%{repository} (%{commit}: %{author}): %{message}"
       - "More info : %{build_url}"

AUTHORS

@@ -6,7 +6,9 @@ This list is sorted alphabetically by first name.
 
 Felix Frank
 James Shubin
+Joe Groocock
 Johan Bloemberg
 Jonathan Gold
 Julien Pivotto
 Paul Morgan
+Samuel Gélineau

COPYING

@@ -1,7 +1,7 @@
                     GNU GENERAL PUBLIC LICENSE
                        Version 3, 29 June 2007
 
- Copyright (C) 2007 Free Software Foundation, Inc. <http://fsf.org/>
+ Copyright (C) 2007 Free Software Foundation, Inc. <https://fsf.org/>
  Everyone is permitted to copy and distribute verbatim copies
  of this license document, but changing it is not allowed.
 
@@ -645,7 +645,7 @@ the "copyright" line and a pointer to where the full notice is found.
     GNU General Public License for more details.
 
     You should have received a copy of the GNU General Public License
-    along with this program.  If not, see <http://www.gnu.org/licenses/>.
+    along with this program.  If not, see <https://www.gnu.org/licenses/>.
 
 Also add information on how to contact you by electronic and paper mail.
 
@@ -664,11 +664,11 @@ might be different; for a GUI interface, you would use an "about box".
   You should also get your employer (if you work as a programmer) or school,
 if any, to sign a "copyright disclaimer" for the program, if necessary.
 For more information on this, and how to apply and follow the GNU GPL, see
-<http://www.gnu.org/licenses/>.
+<https://www.gnu.org/licenses/>.
 
   The GNU General Public License does not permit incorporating your program
 into proprietary programs.  If your program is a subroutine library, you
 may consider it more useful to permit linking proprietary applications with
 the library.  If this is what you want to do, use the GNU Lesser General
 Public License instead of this License.  But first, please read
-<http://www.gnu.org/philosophy/why-not-lgpl.html>.
+<https://www.gnu.org/philosophy/why-not-lgpl.html>.

COPYRIGHT

@@ -1,5 +1,5 @@
 Mgmt
-Copyright (C) 2013-2018+ James Shubin and the project contributors
+Copyright (C) 2013-2024+ James Shubin and the project contributors
 Written by James Shubin <james@shubin.ca> and the project contributors
 
 This program is free software: you can redistribute it and/or modify
@@ -13,4 +13,16 @@ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 GNU General Public License for more details.
 
 You should have received a copy of the GNU General Public License
-along with this program.  If not, see <http://www.gnu.org/licenses/>.
+along with this program.  If not, see <https://www.gnu.org/licenses/>.
+
+Additional permission under GNU GPL version 3 section 7
+
+If you modify this program, or any covered work, by linking or combining it
+with embedded mcl code and modules (and that the embedded mcl code and
+modules which link with this program, contain a copy of their source code in
+the authoritative form) containing parts covered by the terms of any other
+license, the licensors of this program grant you additional permission to
+convey the resulting work. Furthermore, the licensors of this program grant
+the original author, James Shubin, additional permission to update this
+additional permission if he deems it necessary to achieve the goals of this
+additional permission.

Makefile

@@ -1,5 +1,5 @@
 # Mgmt
-# Copyright (C) 2013-2018+ James Shubin and the project contributors
+# Copyright (C) 2013-2024+ James Shubin and the project contributors
 # Written by James Shubin <james@shubin.ca> and the project contributors
 #
 # This program is free software: you can redistribute it and/or modify
@@ -13,14 +13,31 @@
 # GNU General Public License for more details.
 #
 # You should have received a copy of the GNU General Public License
-# along with this program.  If not, see <http://www.gnu.org/licenses/>.
+# along with this program.  If not, see <https://www.gnu.org/licenses/>.
+#
+# Additional permission under GNU GPL version 3 section 7
+#
+# If you modify this program, or any covered work, by linking or combining it
+# with embedded mcl code and modules (and that the embedded mcl code and
+# modules which link with this program, contain a copy of their source code in
+# the authoritative form) containing parts covered by the terms of any other
+# license, the licensors of this program grant you additional permission to
+# convey the resulting work. Furthermore, the licensors of this program grant
+# the original author, James Shubin, additional permission to update this
+# additional permission if he deems it necessary to achieve the goals of this
+# additional permission.
 
 SHELL = /usr/bin/env bash
-.PHONY: all art cleanart version program lang path deps run race bindata generate build build-debug crossbuild clean test gofmt yamlfmt format docs rpmbuild mkdirs rpm srpm spec tar upload upload-sources upload-srpms upload-rpms copr tag release
-.SILENT: clean bindata
+.PHONY: all art cleanart version program lang path deps run race generate build build-debug crossbuild clean test gofmt yamlfmt format docs
+.PHONY: rpmbuild mkdirs rpm srpm spec tar upload upload-sources upload-srpms upload-rpms upload-releases copr tag
+.PHONY: mkosi mkosi_fedora-latest mkosi_fedora-older mkosi_stream-latest mkosi_debian-stable mkosi_ubuntu-latest mkosi_archlinux
+.PHONY: release release_test releases_path release_binary_amd64 release_binary_arm64 release_fedora-latest release_fedora-older release_stream-latest release_debian-stable release_ubuntu-latest release_archlinux
+.PHONY: funcgen
+.SILENT: clean
 
 # a large amount of output from this `find`, can cause `make` to be much slower!
 GO_FILES := $(shell find * -name '*.go' -not -path 'old/*' -not -path 'tmp/*')
+MCL_FILES := $(shell find lang/ -name '*.mcl' -not -path 'old/*' -not -path 'tmp/*')
 
 SVERSION := $(or $(SVERSION),$(shell git describe --match '[0-9]*\.[0-9]*\.[0-9]*' --tags --dirty --always))
 VERSION := $(or $(VERSION),$(shell git describe --match '[0-9]*\.[0-9]*\.[0-9]*' --tags --abbrev=0))
@@ -31,6 +48,13 @@ ifeq ($(VERSION),$(SVERSION))
 else
 	RELEASE = untagged
 endif
+# debugging is harder if -trimpath is set, so disable it if env var is set
+# this is force enabled automatically when using the release target
+ifeq ($(MGMT_NOTRIMPATH),true)
+	TRIMPATH =
+else
+	TRIMPATH = -trimpath
+endif
 ARCH = $(uname -m)
 SPEC = rpmbuild/SPECS/$(PROGRAM).spec
 SOURCE = rpmbuild/SOURCES/$(PROGRAM)-$(VERSION).tar.bz2
@@ -48,9 +72,68 @@ GOOSARCHES ?= linux/amd64 linux/ppc64 linux/ppc64le linux/arm64 darwin/amd64
 GOHOSTOS = $(shell go env GOHOSTOS)
 GOHOSTARCH = $(shell go env GOHOSTARCH)
 
-RPM_PKG = releases/$(VERSION)/rpm/mgmt-$(VERSION)-1.x86_64.rpm
-DEB_PKG = releases/$(VERSION)/deb/mgmt_$(VERSION)_amd64.deb
-PACMAN_PKG = releases/$(VERSION)/pacman/mgmt-$(VERSION)-1-x86_64.pkg.tar.xz
+# The underscores separate the prefix name ("TOKEN") the distro ("BINARY",
+# "FEDORA-LATEST", etc...) and the arch ("AMD64"). The distro can have a dash.
+TOKEN_BINARY_AMD64 = $(shell grep -v '#' releases/binary-amd64.release)
+TOKEN_BINARY_ARM64 = $(shell grep -v '#' releases/binary-arm64.release)
+TOKEN_FEDORA-LATEST = $(shell grep -v '#' releases/fedora-latest.release)
+TOKEN_FEDORA-OLDER = $(shell grep -v '#' releases/fedora-older.release)
+TOKEN_STREAM-LATEST = $(shell grep -v '#' releases/stream-latest.release)
+TOKEN_DEBIAN-STABLE = $(shell grep -v '#' releases/debian-stable.release)
+TOKEN_UBUNTU-LATEST = $(shell grep -v '#' releases/ubuntu-latest.release)
+TOKEN_ARCHLINUX = $(shell grep -v '#' releases/archlinux.release)
+
+FILE_BINARY_AMD64 = mgmt-linux-amd64-$(VERSION)
+FILE_BINARY_ARM64 = mgmt-linux-arm64-$(VERSION)
+# TODO: add ARCH onto these at the end, eg _AMD64
+FILE_FEDORA-LATEST = mgmt-$(TOKEN_FEDORA-LATEST)-$(VERSION)-1.x86_64.rpm
+FILE_FEDORA-OLDER = mgmt-$(TOKEN_FEDORA-OLDER)-$(VERSION)-1.x86_64.rpm
+FILE_STREAM-LATEST = mgmt-$(TOKEN_STREAM-LATEST)-$(VERSION)-1.x86_64.rpm
+FILE_DEBIAN-STABLE = mgmt_$(TOKEN_DEBIAN-STABLE)_$(VERSION)_amd64.deb
+FILE_UBUNTU-LATEST = mgmt_$(TOKEN_UBUNTU-LATEST)_$(VERSION)_amd64.deb
+FILE_ARCHLINUX = mgmt-$(TOKEN_ARCHLINUX)-$(VERSION)-1-x86_64.pkg.tar.xz
+
+PKG_BINARY_AMD64 = releases/$(VERSION)/$(TOKEN_BINARY_AMD64)/$(FILE_BINARY_AMD64)
+PKG_BINARY_ARM64 = releases/$(VERSION)/$(TOKEN_BINARY_ARM64)/$(FILE_BINARY_ARM64)
+PKG_FEDORA-LATEST = releases/$(VERSION)/$(TOKEN_FEDORA-LATEST)/$(FILE_FEDORA-LATEST)
+PKG_FEDORA-OLDER = releases/$(VERSION)/$(TOKEN_FEDORA-OLDER)/$(FILE_FEDORA-OLDER)
+PKG_STREAM-LATEST = releases/$(VERSION)/$(TOKEN_STREAM-LATEST)/$(FILE_STREAM-LATEST)
+PKG_DEBIAN-STABLE = releases/$(VERSION)/$(TOKEN_DEBIAN-STABLE)/$(FILE_DEBIAN-STABLE)
+PKG_UBUNTU-LATEST = releases/$(VERSION)/$(TOKEN_UBUNTU-LATEST)/$(FILE_UBUNTU-LATEST)
+PKG_ARCHLINUX = releases/$(VERSION)/$(TOKEN_ARCHLINUX)/$(FILE_ARCHLINUX)
+
+DEP_BINARY_AMD64 =
+ifneq ($(TOKEN_BINARY_AMD64),)
+	DEP_BINARY_AMD64 = $(PKG_BINARY_AMD64)
+endif
+DEP_BINARY_ARM64 =
+ifneq ($(TOKEN_BINARY_ARM64),)
+	DEP_BINARY_ARM64 = $(PKG_BINARY_ARM64)
+endif
+DEP_FEDORA-LATEST =
+ifneq ($(TOKEN_FEDORA-LATEST),)
+	DEP_FEDORA-LATEST = $(PKG_FEDORA-LATEST)
+endif
+DEP_FEDORA-OLDER =
+ifneq ($(TOKEN_FEDORA-OLDER),)
+	DEP_FEDORA-OLDER = $(PKG_FEDORA-OLDER)
+endif
+DEP_STREAM-LATEST =
+ifneq ($(TOKEN_STREAM-LATEST),)
+	DEP_STREAM-LATEST = $(PKG_STREAM-LATEST)
+endif
+DEP_DEBIAN-STABLE =
+ifneq ($(TOKEN_DEBIAN-STABLE),)
+	DEP_DEBIAN-STABLE = $(PKG_DEBIAN-STABLE)
+endif
+DEP_UBUNTU-LATEST =
+ifneq ($(TOKEN_UBUNTU-LATEST),)
+	DEP_UBUNTU-LATEST = $(PKG_UBUNTU-LATEST)
+endif
+DEP_ARCHLINUX =
+ifneq ($(TOKEN_ARCHLINUX),)
+	DEP_ARCHLINUX = $(PKG_ARCHLINUX)
+endif
 
 SHA256SUMS = releases/$(VERSION)/SHA256SUMS
 SHA256SUMS_ASC = $(SHA256SUMS).asc
@@ -115,27 +198,21 @@ run: ## run mgmt
 race:
 	find . -maxdepth 1 -type f -name '*.go' -not -name '*_test.go' | xargs go run -race -ldflags "-X main.program=$(PROGRAM) -X main.version=$(SVERSION)"
 
-# generate go files from non-go source
-bindata: ## generate go files from non-go sources
-	@echo "Generating: bindata..."
-	$(MAKE) --quiet -C bindata
-
 generate:
 	go generate
 
 lang: ## generates the lexer/parser for the language frontend
 	@# recursively run make in child dir named lang
-	@echo "Generating: lang..."
-	$(MAKE) --quiet -C lang
+	@$(MAKE) --quiet -C lang
 
 # build a `mgmt` binary for current host os/arch
 $(PROGRAM): build/mgmt-${GOHOSTOS}-${GOHOSTARCH} ## build an mgmt binary for current host os/arch
 	cp -a $< $@
 
-$(PROGRAM).static: $(GO_FILES)
+$(PROGRAM).static: $(GO_FILES) $(MCL_FILES) go.mod go.sum
 	@echo "Building: $(PROGRAM).static, version: $(SVERSION)..."
 	go generate
-	go build -a -installsuffix cgo -tags netgo -ldflags '-extldflags "-static" -X main.program=$(PROGRAM) -X main.version=$(SVERSION) -s -w' -o $(PROGRAM).static $(BUILD_FLAGS);
+	go build $(TRIMPATH) -a -installsuffix cgo -tags netgo -ldflags '-extldflags "-static" -X main.program=$(PROGRAM) -X main.version=$(SVERSION) -s -w' -o $(PROGRAM).static $(BUILD_FLAGS);
 
 build: LDFLAGS=-s -w ## build a fresh mgmt binary
 build: $(PROGRAM)
@@ -147,23 +224,23 @@ build-debug: $(PROGRAM)
 # extract os and arch from target pattern
 GOOS=$(firstword $(subst -, ,$*))
 GOARCH=$(lastword $(subst -, ,$*))
-build/mgmt-%: $(GO_FILES) | bindata lang
+build/mgmt-%: $(GO_FILES) $(MCL_FILES) go.mod go.sum | lang funcgen
 	@echo "Building: $(PROGRAM), os/arch: $*, version: $(SVERSION)..."
-	@# reassigning GOOS and GOARCH to make build command copy/pastable
-	@# go 1.10 requires specifying the package for ldflags
-	@if go version | grep -qE 'go1.9'; then \
-		time env GOOS=${GOOS} GOARCH=${GOARCH} go build -i -ldflags "-X main.program=$(PROGRAM) -X main.version=$(SVERSION) ${LDFLAGS}" -o $@ $(BUILD_FLAGS); \
-	else \
-		time env GOOS=${GOOS} GOARCH=${GOARCH} go build -i -ldflags=$(PKGNAME)="-X main.program=$(PROGRAM) -X main.version=$(SVERSION) ${LDFLAGS}" -o $@ $(BUILD_FLAGS); \
-	fi
+	time env GOOS=${GOOS} GOARCH=${GOARCH} go build $(TRIMPATH) -ldflags=$(PKGNAME)="-X main.program=$(PROGRAM) -X main.version=$(SVERSION) ${LDFLAGS}" -o $@ $(BUILD_FLAGS)
 
 # create a list of binary file names to use as make targets
+# to use this you might want to run something like:
+# GOOSARCHES='linux/arm64' GOTAGS='noaugeas novirt' make crossbuild
+# and the output will end up in build/
 crossbuild_targets = $(addprefix build/mgmt-,$(subst /,-,${GOOSARCHES}))
 crossbuild: ${crossbuild_targets}
 
 clean: ## clean things up
-	$(MAKE) --quiet -C bindata clean
+	$(MAKE) --quiet -C test clean
 	$(MAKE) --quiet -C lang clean
+	$(MAKE) --quiet -C misc/mkosi clean
+	rm -f lang/core/generated_funcs.go || true
+	rm -f lang/core/generated_funcs_test.go || true
 	[ ! -e $(PROGRAM) ] || rm $(PROGRAM)
 	rm -f *_stringer.go	# generated by `go generate`
 	rm -f *_mock.go		# generated by `go generate`
@@ -171,6 +248,8 @@ clean: ## clean things up
 	rm -f build/mgmt-*
 
 test: build ## run tests
+	@# recursively run make in child dir named test
+	@$(MAKE) --quiet -C test
 	./test.sh
 
 # create all test targets for make tab completion (eg: make test-gofmt)
@@ -186,8 +265,8 @@ $(addprefix test-shell-,${test_shell}): test-shell-%: build
 
 gofmt:
 	# TODO: remove gofmt once goimports has a -s option
-	find . -maxdepth 6 -type f -name '*.go' -not -path './old/*' -not -path './tmp/*' -not -path './vendor/*' -exec gofmt -s -w {} \;
-	find . -maxdepth 6 -type f -name '*.go' -not -path './old/*' -not -path './tmp/*' -not -path './vendor/*' -exec goimports -w {} \;
+	find . -maxdepth 9 -type f -name '*.go' -not -path './old/*' -not -path './tmp/*' -not -path './vendor/*' -exec gofmt -s -w {} \;
+	find . -maxdepth 9 -type f -name '*.go' -not -path './old/*' -not -path './tmp/*' -not -path './vendor/*' -exec goimports -w {} \;
 
 yamlfmt:
 	find . -maxdepth 3 -type f -name '*.yaml' -not -path './old/*' -not -path './tmp/*' -not -path './omv.yaml' -exec ruby -e "require 'yaml'; x=YAML.load_file('{}').to_yaml.each_line.map(&:rstrip).join(10.chr)+10.chr; File.open('{}', 'w').write x" \;
@@ -324,6 +403,10 @@ upload-rpms: rpmbuild/RPMS/ rpmbuild/RPMS/SHA256SUMS rpmbuild/RPMS/SHA256SUMS.as
 		rsync -avz --prune-empty-dirs rpmbuild/RPMS/ $(SERVER):$(REMOTE_PATH)/RPMS/; \
 	fi
 
+upload-releases:
+	echo Running releases/ upload...
+	rsync -avz --exclude '.mkdir' --exclude 'mgmt-release.url' releases/ $(SERVER):$(REMOTE_PATH)/releases/
+
 #
 #	copr build
 #
@@ -336,18 +419,105 @@ copr: upload-srpms ## build in copr
 tag: ## tags a new release
 	./misc/tag.sh
 
+#
+#	mkosi
+#
+mkosi: mkosi_fedora-latest mkosi_fedora-older mkosi_stream-latest mkosi_debian-stable mkosi_ubuntu-latest mkosi_archlinux ## builds distro packages via mkosi
+
+mkosi_fedora-latest: releases/$(VERSION)/.mkdir
+	@title='$@' ; echo "Generating: $${title#'mkosi_'} via mkosi..."
+	@title='$@' ; distro=$${title#'mkosi_'} ; ./misc/mkosi/make.sh $${distro} `realpath "releases/$(VERSION)/"`
+
+mkosi_fedora-older: releases/$(VERSION)/.mkdir
+	@title='$@' ; echo "Generating: $${title#'mkosi_'} via mkosi..."
+	@title='$@' ; distro=$${title#'mkosi_'} ; ./misc/mkosi/make.sh $${distro} `realpath "releases/$(VERSION)/"`
+
+mkosi_stream-latest: releases/$(VERSION)/.mkdir
+	@title='$@' ; echo "Generating: $${title#'mkosi_'} via mkosi..."
+	@title='$@' ; distro=$${title#'mkosi_'} ; ./misc/mkosi/make.sh $${distro} `realpath "releases/$(VERSION)/"`
+
+mkosi_debian-stable: releases/$(VERSION)/.mkdir
+	@title='$@' ; echo "Generating: $${title#'mkosi_'} via mkosi..."
+	@title='$@' ; distro=$${title#'mkosi_'} ; ./misc/mkosi/make.sh $${distro} `realpath "releases/$(VERSION)/"`
+
+mkosi_ubuntu-latest: releases/$(VERSION)/.mkdir
+	@title='$@' ; echo "Generating: $${title#'mkosi_'} via mkosi..."
+	@title='$@' ; distro=$${title#'mkosi_'} ; ./misc/mkosi/make.sh $${distro} `realpath "releases/$(VERSION)/"`
+
+mkosi_archlinux: releases/$(VERSION)/.mkdir
+	@title='$@' ; echo "Generating: $${title#'mkosi_'} via mkosi..."
+	@title='$@' ; distro=$${title#'mkosi_'} ; ./misc/mkosi/make.sh $${distro} `realpath "releases/$(VERSION)/"`
+
 #
 #	release
 #
+release: TRIMPATH = -trimpath
 release: releases/$(VERSION)/mgmt-release.url ## generates and uploads a release
 
-releases/$(VERSION)/mgmt-release.url: $(RPM_PKG) $(DEB_PKG) $(PACMAN_PKG) $(SHA256SUMS_ASC)
+releases_path:
+	@#Don't put any other output or dependencies in here or they'll show!
+	@echo "releases/$(VERSION)/"
+
+release_test: $(DEP_BINARY_AMD64) $(DEP_BINARY_ARM64) $(DEP_FEDORA-LATEST) $(DEP_FEDORA-OLDER) $(DEP_STREAM-LATEST) $(DEP_DEBIAN-STABLE) $(DEP_UBUNTU-LATEST) $(DEP_ARCHLINUX) $(SHA256SUMS_ASC)
+	@echo '$$< denotes ‘the first dependency of the current rule’.'
+	@echo '> '"$<"
+	@echo
+	@echo '$$@ denotes ‘the target of the current rule’.'
+	@echo '> '"$@"
+	@echo
+	@echo '$$^ denotes ‘the dependencies of the current rule’.'
+	@echo '> '"$^"
+	@echo
+	@echo '$$* denotes ‘the stem with which the pattern of the current rule matched’.'
+	@echo '> '"$*"
+	@echo
+	@echo "TOKEN_BINARY_AMD64: $(TOKEN_BINARY_AMD64)"
+	@echo "DEP_BINARY_AMD64: $(DEP_BINARY_AMD64)"
+	@echo
+	@echo "TOKEN_BINARY_ARM64: $(TOKEN_BINARY_ARM64)"
+	@echo "DEP_BINARY_ARM64: $(DEP_BINARY_ARM64)"
+	@echo
+	@echo "TOKEN_FEDORA-LATEST: $(TOKEN_FEDORA-LATEST)"
+	@echo "DEP_FEDORA-LATEST: $(DEP_FEDORA-LATEST)"
+	@echo
+	@echo "TOKEN_FEDORA-OLDER: $(TOKEN_FEDORA-OLDER)"
+	@echo "DEP_FEDORA-OLDER: $(DEP_FEDORA-OLDER)"
+	@echo
+	@echo "TOKEN_STREAM-LATEST: $(TOKEN_STREAM-LATEST)"
+	@echo "DEP_STREAM-LATEST: $(DEP_STREAM-LATEST)"
+	@echo
+	@echo "TOKEN_DEBIAN-STABLE: $(TOKEN_DEBIAN-STABLE)"
+	@echo "DEP_DEBIAN-STABLE: $(DEP_DEBIAN-STABLE)"
+	@echo
+	@echo "TOKEN_UBUNTU-LATEST: $(TOKEN_UBUNTU-LATEST)"
+	@echo "DEP_UBUNTU-LATEST: $(DEP_UBUNTU-LATEST)"
+	@echo
+	@echo "TOKEN_ARCHLINUX: $(TOKEN_ARCHLINUX)"
+	@echo "DEP_ARCHLINUX: $(DEP_ARCHLINUX)"
+
+release_binary_amd64: $(PKG_BINARY_AMD64)
+release_binary_arm64: $(PKG_BINARY_ARM64)
+release_fedora-latest: $(PKG_FEDORA-LATEST)
+release_fedora-older: $(PKG_FEDORA-OLDER)
+release_stream-latest: $(PKG_STREAM-LATEST)
+release_debian-stable: $(PKG_DEBIAN-STABLE)
+release_ubuntu-latest: $(PKG_UBUNTU-LATEST)
+release_archlinux: $(PKG_ARCHLINUX)
+
+releases/$(VERSION)/mgmt-release.url: $(DEP_BINARY_AMD64) $(DEP_BINARY_ARM64) $(DEP_FEDORA-LATEST) $(DEP_FEDORA-OLDER) $(DEP_STREAM-LATEST) $(DEP_DEBIAN-STABLE) $(DEP_UBUNTU-LATEST) $(DEP_ARCHLINUX) $(SHA256SUMS_ASC)
+	@echo "Pushing git tag $(VERSION) to origin..."
+	git push origin $(VERSION)
 	@echo "Creating github release..."
 	hub release create \
 		-F <( echo -e "$(VERSION)\n";echo "Verify the signatures of all packages before you use them. The signing key can be downloaded from https://purpleidea.com/contact/#pgp-key to verify the release." ) \
-		-a $(RPM_PKG) \
-		-a $(DEB_PKG) \
-		-a $(PACMAN_PKG) \
+		` [ -e $(PKG_BINARY_AMD64) ] && printf -- "-a $(PKG_BINARY_AMD64)" ` \
+		` [ -e $(PKG_BINARY_ARM64) ] && printf -- "-a $(PKG_BINARY_ARM64)" ` \
+		` [ -e $(PKG_FEDORA-LATEST) ] && printf -- "-a $(PKG_FEDORA-LATEST)" ` \
+		` [ -e $(PKG_FEDORA-OLDER) ] && printf -- "-a $(PKG_FEDORA-OLDER)" ` \
+		` [ -e $(PKG_STREAM-LATEST) ] && printf -- "-a $(PKG_STREAM-LATEST)" ` \
+		` [ -e $(PKG_DEBIAN-STABLE) ] && printf -- "-a $(PKG_DEBIAN-STABLE)" ` \
+		` [ -e $(PKG_UBUNTU-LATEST) ] && printf -- "-a $(PKG_UBUNTU-LATEST)" ` \
+		` [ -e $(PKG_ARCHLINUX) ] && printf -- "-a $(PKG_ARCHLINUX)" ` \
 		-a $(SHA256SUMS_ASC) \
 		$(VERSION) \
 		> releases/$(VERSION)/mgmt-release.url \
@@ -355,32 +525,92 @@ releases/$(VERSION)/mgmt-release.url: $(RPM_PKG) $(DEB_PKG) $(PACMAN_PKG) $(SHA2
 		|| rm -f releases/$(VERSION)/mgmt-release.url
 
 releases/$(VERSION)/.mkdir:
-	mkdir -p releases/$(VERSION)/{deb,rpm,pacman}/ && touch releases/$(VERSION)/.mkdir
+	mkdir -p \
+	` [ "$(TOKEN_BINARY_AMD64)" != "" ] && printf -- "releases/$(VERSION)/$(TOKEN_BINARY_AMD64)/" ` \
+	` [ "$(TOKEN_BINARY_ARM64)" != "" ] && printf -- "releases/$(VERSION)/$(TOKEN_BINARY_ARM64)/" ` \
+	` [ "$(TOKEN_FEDORA-LATEST)" != "" ] && printf -- "releases/$(VERSION)/$(TOKEN_FEDORA-LATEST)/" ` \
+	` [ "$(TOKEN_FEDORA-OLDER)" != "" ] && printf -- "releases/$(VERSION)/$(TOKEN_FEDORA-OLDER)/" ` \
+	` [ "$(TOKEN_STREAM-LATEST)" != "" ] && printf -- "releases/$(VERSION)/$(TOKEN_STREAM-LATEST)/" ` \
+	` [ "$(TOKEN_DEBIAN-STABLE)" != "" ] && printf -- "releases/$(VERSION)/$(TOKEN_DEBIAN-STABLE)/" ` \
+	` [ "$(TOKEN_UBUNTU-LATEST)" != "" ] && printf -- "releases/$(VERSION)/$(TOKEN_UBUNTU-LATEST)/" ` \
+	` [ "$(TOKEN_ARCHLINUX)" != "" ] && printf -- "releases/$(VERSION)/$(TOKEN_ARCHLINUX)/" ` \
+	&& touch releases/$(VERSION)/.mkdir
 
-releases/$(VERSION)/rpm/changelog: $(PROGRAM) releases/$(VERSION)/.mkdir
-	@echo "Generating rpm changelog..."
-	./misc/make-rpm-changelog.sh $(VERSION)
+# These are defined conditionally, since if the token is empty, they warn!
+ifneq ($(TOKEN_BINARY_AMD64),)
+$(PKG_BINARY_AMD64): build/mgmt-linux-amd64 releases/$(VERSION)/.mkdir
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; echo "Building: $${distro} package..."
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; cp -a build/mgmt-linux-amd64 $(PKG_BINARY_AMD64)
+endif
+ifneq ($(TOKEN_BINARY_ARM64),)
+$(PKG_BINARY_ARM64): build/mgmt-linux-arm64 releases/$(VERSION)/.mkdir
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; echo "Building: $${distro} package..."
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; cp -a build/mgmt-linux-arm64 $(PKG_BINARY_ARM64)
+endif
+ifneq ($(TOKEN_FEDORA-LATEST),)
+releases/$(VERSION)/$(TOKEN_FEDORA-LATEST)/changelog: $(PROGRAM) releases/$(VERSION)/.mkdir
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; echo "Generating: $${distro} changelog..."
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; ./misc/make-rpm-changelog.sh "$${distro}" $(VERSION)
 
-$(RPM_PKG): releases/$(VERSION)/rpm/changelog
-	@echo "Building rpm package..."
-	./misc/fpm-pack.sh rpm $(VERSION) libvirt-devel augeas-devel
+$(PKG_FEDORA-LATEST): releases/$(VERSION)/$(TOKEN_FEDORA-LATEST)/changelog
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; echo "Building: $${distro} package..."
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; ./misc/fpm-pack.sh $${distro} $(VERSION) "$(FILE_FEDORA-LATEST)" libvirt-devel augeas-devel
+endif
+ifneq ($(TOKEN_FEDORA-OLDER),)
+releases/$(VERSION)/$(TOKEN_FEDORA-OLDER)/changelog: $(PROGRAM) releases/$(VERSION)/.mkdir
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; echo "Generating: $${distro} changelog..."
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; ./misc/make-rpm-changelog.sh "$${distro}" $(VERSION)
 
-releases/$(VERSION)/deb/changelog: $(PROGRAM) releases/$(VERSION)/.mkdir
-	@echo "Generating deb changelog..."
-	./misc/make-deb-changelog.sh $(VERSION)
+$(PKG_FEDORA-OLDER): releases/$(VERSION)/$(TOKEN_FEDORA-OLDER)/changelog
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; echo "Building: $${distro} package..."
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; ./misc/fpm-pack.sh $${distro} $(VERSION) "$(FILE_FEDORA-OLDER)" libvirt-devel augeas-devel
+endif
+ifneq ($(TOKEN_STREAM-LATEST),)
+releases/$(VERSION)/$(TOKEN_STREAM-LATEST)/changelog: $(PROGRAM) releases/$(VERSION)/.mkdir
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; echo "Generating: $${distro} changelog..."
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; ./misc/make-rpm-changelog.sh "$${distro}" $(VERSION)
 
-$(DEB_PKG): releases/$(VERSION)/deb/changelog
-	@echo "Building deb package..."
-	./misc/fpm-pack.sh deb $(VERSION) libvirt-dev libaugeas-dev
+$(PKG_STREAM-LATEST): releases/$(VERSION)/$(TOKEN_STREAM-LATEST)/changelog
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; echo "Building: $${distro} package..."
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; ./misc/fpm-pack.sh $${distro} $(VERSION) "$(FILE_STREAM-LATEST)" libvirt-devel augeas-devel
+endif
+ifneq ($(TOKEN_DEBIAN-STABLE),)
+releases/$(VERSION)/$(TOKEN_DEBIAN-STABLE)/changelog: $(PROGRAM) releases/$(VERSION)/.mkdir
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; echo "Generating: $${distro} changelog..."
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; ./misc/make-deb-changelog.sh "$${distro}" $(VERSION)
 
-$(PACMAN_PKG): $(PROGRAM) releases/$(VERSION)/.mkdir
-	@echo "Building pacman package..."
-	./misc/fpm-pack.sh pacman $(VERSION) libvirt augeas
+$(PKG_DEBIAN-STABLE): releases/$(VERSION)/$(TOKEN_DEBIAN-STABLE)/changelog
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; echo "Building: $${distro} package..."
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; ./misc/fpm-pack.sh $${distro} $(VERSION) "$(FILE_DEBIAN-STABLE)" libvirt-dev libaugeas-dev
+endif
+ifneq ($(TOKEN_UBUNTU-LATEST),)
+releases/$(VERSION)/$(TOKEN_UBUNTU-LATEST)/changelog: $(PROGRAM) releases/$(VERSION)/.mkdir
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; echo "Generating: $${distro} changelog..."
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; ./misc/make-deb-changelog.sh "$${distro}" $(VERSION)
 
-$(SHA256SUMS): $(RPM_PKG) $(DEB_PKG) $(PACMAN_PKG)
+$(PKG_UBUNTU-LATEST): releases/$(VERSION)/$(TOKEN_UBUNTU-LATEST)/changelog
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; echo "Building: $${distro} package..."
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; ./misc/fpm-pack.sh $${distro} $(VERSION) "$(FILE_UBUNTU-LATEST)" libvirt-dev libaugeas-dev
+endif
+ifneq ($(TOKEN_ARCHLINUX),)
+$(PKG_ARCHLINUX): $(PROGRAM) releases/$(VERSION)/.mkdir
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; echo "Building: $${distro} package..."
+	@title='$(@D)' ; distro=$${title#'releases/$(VERSION)/'} ; ./misc/fpm-pack.sh $${distro} $(VERSION) "$(FILE_ARCHLINUX)" libvirt augeas
+endif
+
+$(SHA256SUMS): $(DEP_BINARY_AMD64) $(DEP_BINARY_ARM64) $(DEP_FEDORA-LATEST) $(DEP_FEDORA-OLDER) $(DEP_STREAM-LATEST) $(DEP_DEBIAN-STABLE) $(DEP_UBUNTU-LATEST) $(DEP_ARCHLINUX)
 	@# remove the directory separator in the SHA256SUMS file
-	@echo "Generating sha256 sum..."
-	sha256sum $(RPM_PKG) $(DEB_PKG) $(PACMAN_PKG) | awk -F '/| ' '{print $$1"  "$$6}' > $(SHA256SUMS)
+	@echo "Generating: sha256 sum..."
+	sha256sum \
+	` [ -e $(PKG_BINARY_AMD64) ] && printf -- "$(PKG_BINARY_AMD64)" ` \
+	` [ -e $(PKG_BINARY_ARM64) ] && printf -- "$(PKG_BINARY_ARM64)" ` \
+	` [ -e $(PKG_FEDORA-LATEST) ] && printf -- "$(PKG_FEDORA-LATEST)" ` \
+	` [ -e $(PKG_FEDORA-OLDER) ] && printf -- "$(PKG_FEDORA-OLDER)" ` \
+	` [ -e $(PKG_STREAM-LATEST) ] && printf -- "$(PKG_STREAM-LATEST)" ` \
+	` [ -e $(PKG_DEBIAN-STABLE) ] && printf -- "$(PKG_DEBIAN-STABLE)" ` \
+	` [ -e $(PKG_UBUNTU-LATEST) ] && printf -- "$(PKG_UBUNTU-LATEST)" ` \
+	` [ -e $(PKG_ARCHLINUX) ] && printf -- "$(PKG_ARCHLINUX)" ` \
+	| awk -F '/| ' '{print $$1"  "$$6}' > $(SHA256SUMS)
 
 $(SHA256SUMS_ASC): $(SHA256SUMS)
 	@echo "Signing sha256 sum..."
@@ -406,4 +636,10 @@ help: ## show this help screen
 		awk 'BEGIN {FS = ":.*?## "}; {printf "\033[36m%-20s\033[0m %s\n", $$1, $$2}'
 	@echo ''
 
+funcgen: lang/core/generated_funcs.go
+
+lang/core/generated_funcs.go: lang/funcs/funcgen/*.go lang/core/funcgen.yaml lang/funcs/funcgen/templates/generated_funcs.go.tpl
+	@echo "Generating: funcs..."
+	@go run `find lang/funcs/funcgen/ -maxdepth 1 -type f -name '*.go' -not -name '*_test.go'` -templates=lang/funcs/funcgen/templates/generated_funcs.go.tpl >/dev/null
+
 # vim: ts=8

README.md

@@ -3,23 +3,74 @@
 [![mgmt!](art/mgmt.png)](art/)
 
 [![Go Report Card](https://goreportcard.com/badge/github.com/purpleidea/mgmt?style=flat-square)](https://goreportcard.com/report/github.com/purpleidea/mgmt)
-[![Build Status](https://img.shields.io/travis/purpleidea/mgmt/master.svg?style=flat-square)](http://travis-ci.org/purpleidea/mgmt)
-[![GoDoc](https://img.shields.io/badge/godoc-reference-5272B4.svg?style=flat-square)](https://godoc.org/github.com/purpleidea/mgmt)
-[![IRC](https://img.shields.io/badge/irc-%23mgmtconfig-orange.svg?style=flat-square)](https://webchat.freenode.net/?channels=#mgmtconfig)
+[![Build Status](https://github.com/purpleidea/mgmt/workflows/.github/workflows/test.yaml/badge.svg)](https://github.com/purpleidea/mgmt/actions/)
+[![GoDoc](https://img.shields.io/badge/godoc-reference-5272B4.svg?style=flat-square)](https://godocs.io/github.com/purpleidea/mgmt)
+[![Matrix](https://img.shields.io/badge/matrix-%23mgmtconfig-orange.svg?style=flat-square)](https://matrix.to/#/#mgmtconfig:matrix.org)
+[![IRC](https://img.shields.io/badge/irc-%23mgmtconfig-orange.svg?style=flat-square)](https://web.libera.chat/?channels=#mgmtconfig)
 [![Patreon](https://img.shields.io/badge/patreon-donate-yellow.svg?style=flat-square)](https://www.patreon.com/purpleidea)
 [![Liberapay](https://img.shields.io/badge/liberapay-donate-yellow.svg?style=flat-square)](https://liberapay.com/purpleidea/donate)
 
+## About:
+
+`Mgmt` is a real-time automation tool. It is familiar to existing configuration
+management software, but is drastically more powerful as it can allow you to
+build real-time, closed-loop feedback systems, in a very safe way, and with a
+surprisingly small amout of our `mcl` code. For example, the following code will
+ensure that your file server is set to read-only when it's friday.
+
+```mcl
+import "datetime"
+$is_friday = datetime.weekday(datetime.now()) == "friday"
+file "/srv/files/" {
+	state => $const.res.file.state.exists,
+	mode => if $is_friday { # this updates the mode, the instant it changes!
+		"0550"
+	} else {
+		"0770"
+	},
+}
+```
+
+It can run continuously, intermittently, or on-demand, and in the first case, it
+will guarantee that your system is always in the desired state for that instant!
+In this mode it can run as a decentralized cluster of agents across your
+network, each exchanging information with the others in real-time, to respond to
+your changing needs. For example, if you want to ensure that some resource runs
+on a maximum of two hosts in your cluster, you can specify that as well:
+
+```mcl
+import "sys"
+import "world"
+
+# we'll set a few scheduling options:
+$opts = struct{strategy => "rr", max => 2, ttl => 10,}
+
+# schedule in a particular namespace with options:
+$set = world.schedule("xsched", $opts)
+
+if sys.hostname() in $set {
+	# use your imagination to put something more complex right here...
+	print "i got scheduled" {} # this will run on the chosen machines
+}
+```
+
+As you add and remove hosts from the cluster, the real-time `schedule` function
+will dynamically pick up to two hosts from the available pool. These specific
+functions aren't intrinsic to the core design, and new ones can be easily added.
+
+Please read on if you'd like to learn more...
+
 ## Community:
 
 Come join us in the `mgmt` community!
 
 | Medium | Link |
 |---|---|
-| IRC | [#mgmtconfig](https://webchat.freenode.net/?channels=#mgmtconfig) on Freenode |
+| Matrix | [#mgmtconfig](https://matrix.to/#/#mgmtconfig:matrix.org) on Matrix.org |
+| IRC | [#mgmtconfig](https://web.libera.chat/?channels=#mgmtconfig) on Libera.Chat |
 | Twitter | [@mgmtconfig](https://twitter.com/mgmtconfig) & [#mgmtconfig](https://twitter.com/hashtag/mgmtconfig) |
-| Mailing list | [mgmtconfig-list@redhat.com](https://www.redhat.com/mailman/listinfo/mgmtconfig-list) |
+| Mailing list | [looking for a new home, suggestions welcome](https://gitlab.freedesktop.org/freedesktop/freedesktop/-/issues/1082) |
 | Patreon | [purpleidea](https://www.patreon.com/purpleidea) on Patreon |
-| Liberapay | [purpleidea](https://liberapay.com/purpleidea/donate) on Liberapay |
 
 ## Status:
 
@@ -30,7 +81,7 @@ approach. The project contains an engine and a language.
 
 Mgmt is a fairly new project. It is usable today, but not yet feature complete.
 With your help you'll be able to influence our design and get us to 1.0 sooner!
-Interested developers should read the [quick start guide](docs/quick-start-guide.md).
+Interested users should read the [quick start guide](docs/quick-start-guide.md).
 
 ## Documentation:
 
@@ -38,7 +89,7 @@ Please read, enjoy and help improve our documentation!
 
 | Documentation | Additional Notes |
 |---|---|
-| [quick start guide](docs/quick-start-guide.md) | for mgmt developers |
+| [quick start guide](docs/quick-start-guide.md) | for everyone |
 | [frequently asked questions](docs/faq.md) | for everyone |
 | [general documentation](docs/documentation.md) | for everyone |
 | [language guide](docs/language-guide.md) | for everyone |
@@ -47,8 +98,9 @@ Please read, enjoy and help improve our documentation!
 | [style guide](docs/style-guide.md) | for mgmt developers |
 | [godoc API reference](https://godoc.org/github.com/purpleidea/mgmt) | for mgmt developers |
 | [prometheus guide](docs/prometheus.md) | for everyone |
-| [puppet guide](docs/puppet-guide.md) | for puppet sysadmins |
 | [development](docs/development.md) | for mgmt developers |
+| [videos](docs/on-the-web.md) | for everyone |
+| [blogs](docs/on-the-web.md) | for everyone |
 
 ## Questions:
 
@@ -57,22 +109,18 @@ If you have a well phrased question that might benefit others, consider asking
 it by sending a patch to the [FAQ](docs/faq.md) section. I'll merge your
 question, and a patch with the answer!
 
-## Roadmap:
+## Get involved:
 
 Feel free to grab one of the straightforward [#mgmtlove](https://github.com/purpleidea/mgmt/labels/mgmtlove)
 issues if you're a first time contributor to the project or if you're unsure
-about what to hack on!
-Please see: [TODO.md](TODO.md) for a list of upcoming work and TODO items.
-Please get involved by working on one of these items or by suggesting something
-else!
+about what to hack on! Please get involved by working on one of these items or
+by suggesting something else! There are some lower priority issues and harder
+issues available in our [TODO](TODO.md) file. Please have a look.
 
 ## Bugs:
 
 Please set the `DEBUG` constant in [main.go](https://github.com/purpleidea/mgmt/blob/master/main.go)
 to `true`, and post the logs when you report the [issue](https://github.com/purpleidea/mgmt/issues).
-Bonus points if you provide a [shell](https://github.com/purpleidea/mgmt/tree/master/test/shell)
-or [OMV](https://github.com/purpleidea/mgmt/tree/master/test/omv) reproducible
-test case.
 Feel free to read my article on [debugging golang programs](https://purpleidea.com/blog/2016/02/15/debugging-golang-programs/).
 
 ## Patches:
@@ -81,6 +129,6 @@ We'd love to have your patches! Please send them by email, or as a pull request.
 
 ## On the web:
 
-[Read what people are saying and publishing about mgmt!](docs/on-the-web.md)
+[Blog posts and recorded talks about mgmt are listed here!](docs/on-the-web.md)
 
 Happy hacking!

TODO.md

@@ -1,10 +1,18 @@
 # TODO
 
-If you're looking for something to do, look here!
-Let us know if you're working on one of the items.
-If you'd like something to work on, ping @purpleidea and I'll create an issue
-tailored especially for you! Just let me know your approximate golang skill
-level and how many hours you'd like to spend on the patch.
+Here is a TODO list of longstanding items that are either lower-priority, or
+more involved in terms of time, skill-level, and/or motivation.
+
+Please have a look, and let us know if you're working on one of the items. It's
+best to open an issue to track your progress and to discuss any implementation
+questions you might have.
+
+Lastly, if you'd like something different to work on, please ping @purpleidea
+and I'll create an issue tailored especially for your approximate golang skill
+level and available time commitment in terms of hours you'd need to spend on the
+patch.
+
+Happy Hacking!
 
 ## Package resource
 
@@ -19,7 +27,7 @@ level and how many hours you'd like to spend on the patch.
 
 ## Svc resource
 
-- [ ] base resource improvements
+- [ ] refreshonly support [:heart:](https://github.com/purpleidea/mgmt/issues/464)
 
 ## Exec resource
 
@@ -33,33 +41,14 @@ level and how many hours you'd like to spend on the patch.
 
 - [ ] automatic edges to file resource [:heart:](https://github.com/purpleidea/mgmt/labels/mgmtlove)
 
-## Virt (libvirt) resource
-
-- [ ] base resource improvements [:heart:](https://github.com/purpleidea/mgmt/labels/mgmtlove)
-
-## Net (systemd-networkd) resource
-
-- [ ] base resource
-
-## Nspawn (systemd-nspawn) resource
-
-- [ ] base resource [:heart:](https://github.com/purpleidea/mgmt/labels/mgmtlove)
-
-## Mount (systemd-mount) resource
-
-- [ ] base resource [:heart:](https://github.com/purpleidea/mgmt/labels/mgmtlove)
-
-## Cron (systemd-timer) resource
-
-- [ ] base resource [:heart:](https://github.com/purpleidea/mgmt/labels/mgmtlove)
-
 ## Http resource
 
 - [ ] base resource [:heart:](https://github.com/purpleidea/mgmt/labels/mgmtlove)
 
 ## Etcd improvements
 
-- [ ] fix embedded etcd master race
+- [ ] fix etcd race bug that only happens during CI testing (intermittently
+failing test case issue)
 
 ## Torrent/dht file transfer
 
@@ -69,17 +58,33 @@ level and how many hours you'd like to spend on the patch.
 
 - [ ] base plumbing
 
+## Resource improvements
+
+- [ ] more reversible resources implemented
+- [ ] more "cloud" resources
+
 ## Language improvements
 
 - [ ] more core functions
 - [ ] automatic language formatter, ala `gofmt`
 - [ ] gedit/gnome-builder/gtksourceview syntax highlighting
 - [ ] vim syntax highlighting
-- [x] emacs syntax highlighting: see `misc/emacs/`
+- [ ] emacs syntax highlighting: see `misc/emacs/` (needs updating)
+- [ ] exposed $error variable for feedback in the language
+- [ ] improve the printf function to add %[]s, %[]f ([]str, []float) and map,
+struct, nested etc... %v would be nice too!
+- [ ] add line/col/file annotations to AST so we can get locations of errors
+that the parser finds
+- [ ] add more error messages with the `%error` pattern in parser.y
+- [ ] we should have helper functions or language sugar to pull a field out of a
+struct, or a value out of a map, or an index out of a list, etc...
+
+## Engine improvements
+
+- [ ] add a "waiting for func" message in the func engine to notify the user
+about slow functions...
 
 ## Other
 
-- [ ] better error/retry handling
-- [ ] deb package target in Makefile
 - [ ] reproducible builds
 - [ ] add your suggestions!

Vagrantfile

@@ -6,7 +6,7 @@ Vagrant.configure(2) do |config|
 	config.vm.synced_folder ".", "/vagrant", disabled: true
 
 	config.vm.define "mgmt-dev" do |instance|
-		instance.vm.box = "fedora/28-cloud-base"
+		instance.vm.box = "bento/fedora-31"
 	end
 
 	config.vm.provider "virtualbox" do |v|
@@ -23,8 +23,7 @@ Vagrant.configure(2) do |config|
 	config.vm.provision "file", source: "vagrant/mgmt.bashrc", destination: ".mgmt.bashrc"
 	config.vm.provision "file", source: "~/.gitconfig", destination: ".gitconfig"
 
-	# copied from make-deps.sh (with added git)
-	config.vm.provision "shell", inline: "dnf install -y libvirt-devel golang golang-googlecode-tools-stringer hg git make gem"
+	config.vm.provision "shell", inline: "dnf install -y golang git make"
 
 	# set up packagekit
 	config.vm.provision "shell" do |shell|
@@ -39,8 +38,10 @@ Vagrant.configure(2) do |config|
 	script = <<-SCRIPT
 		grep -q 'mgmt\.bashrc' ~/.bashrc || echo '. ~/.mgmt.bashrc' >>~/.bashrc
 		. ~/.mgmt.bashrc
-		go get -u github.com/purpleidea/mgmt
-		cd ~/gopath/src/github.com/purpleidea/mgmt
+		mkdir -p ~/gopath/src/github.com/purpleidea
+		cd ~/gopath/src/github.com/purpleidea
+		git clone https://github.com/purpleidea/mgmt --recursive
+		cd mgmt
 		make deps
 	SCRIPT
 	config.vm.provision "shell" do |shell|

bindata/Makefile

@@ -1,38 +0,0 @@
-# Mgmt
-# Copyright (C) 2013-2018+ James Shubin and the project contributors
-# Written by James Shubin <james@shubin.ca> and the project contributors
-#
-# This program is free software: you can redistribute it and/or modify
-# it under the terms of the GNU General Public License as published by
-# the Free Software Foundation, either version 3 of the License, or
-# (at your option) any later version.
-#
-# This program is distributed in the hope that it will be useful,
-# but WITHOUT ANY WARRANTY; without even the implied warranty of
-# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-# GNU General Public License for more details.
-#
-# You should have received a copy of the GNU General Public License
-# along with this program.  If not, see <http://www.gnu.org/licenses/>.
-
-# The bindata target generates go files from any source defined below. To use
-# the files, import the "bindata" package and use:
-# `bytes, err := bindata.Asset("FILEPATH")`
-# where FILEPATH is the path of the original input file relative to `bindata/`.
-
-.PHONY: build clean
-default: build
-
-build: bindata.go
-
-# add more input files as dependencies at the end here...
-bindata.go: ../COPYING
-	# go-bindata --pkg bindata -o <OUTPUT> <INPUT>
-	go-bindata --pkg bindata -o ./$@ $^
-	# gofmt the output file
-	gofmt -s -w $@
-	@ROOT=$$(dirname "$${BASH_SOURCE}")/.. && $$ROOT/misc/header.sh '$@'
-
-clean:
-	# remove generated bindata/*.go
-	@ROOT=$$(dirname "$${BASH_SOURCE}")/.. && rm -f *.go

cli/cli.go

@@ -0,0 +1,165 @@
+// Mgmt
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
+
+// Package cli handles all of the core command line parsing. It's the first
+// entry point after the real main function, and it imports and runs our core
+// "lib".
+package cli
+
+import (
+	"context"
+	"fmt"
+	"os"
+
+	cliUtil "github.com/purpleidea/mgmt/cli/util"
+	"github.com/purpleidea/mgmt/util/errwrap"
+
+	"github.com/alexflint/go-arg"
+)
+
+func init() {
+	if _, err := arg.NewParser(arg.Config{}, &Args{}); err != nil { // sanity check
+		panic(errwrap.Wrapf(err, "invalid args cli struct"))
+	}
+}
+
+// CLI is the entry point for using mgmt normally from the CLI.
+func CLI(ctx context.Context, data *cliUtil.Data) error {
+	// test for sanity
+	if data == nil {
+		return fmt.Errorf("this CLI was not run correctly")
+	}
+	if data.Program == "" || data.Version == "" {
+		return fmt.Errorf("program was not compiled correctly")
+	}
+	if data.Copying == "" {
+		return fmt.Errorf("program copyrights were removed, can't run")
+	}
+
+	args := Args{}
+	args.version = data.Version // copy this in
+	args.description = data.Tagline
+
+	config := arg.Config{
+		Program: data.Program,
+	}
+	parser, err := arg.NewParser(config, &args)
+	if err != nil {
+		// programming error
+		return errwrap.Wrapf(err, "cli config error")
+	}
+	err = parser.Parse(data.Args[1:]) // XXX: args[0] needs to be dropped
+	if err == arg.ErrHelp {
+		parser.WriteHelp(os.Stdout)
+		return nil
+	}
+	if err == arg.ErrVersion {
+		fmt.Printf("%s\n", data.Version) // byon: bring your own newline
+		return nil
+	}
+	if err != nil {
+		//parser.WriteHelp(os.Stdout) // TODO: is doing this helpful?
+		return cliUtil.CliParseError(err) // consistent errors
+	}
+
+	// display the license
+	if args.License {
+		fmt.Printf("%s", data.Copying) // file comes with a trailing nl
+		return nil
+	}
+
+	if ok, err := args.Run(ctx, data); err != nil {
+		return err
+	} else if ok { // did we activate one of the commands?
+		return nil
+	}
+
+	// print help if no subcommands are set
+	parser.WriteHelp(os.Stdout)
+
+	return nil
+}
+
+// Args is the CLI parsing structure and type of the parsed result. This
+// particular struct is the top-most one.
+type Args struct {
+	// XXX: We cannot have both subcommands and a positional argument.
+	// XXX: I think it's a bug of this library that it can't handle argv[0].
+	//Argv0 string `arg:"positional"`
+
+	License bool `arg:"--license" help:"display the license and exit"`
+
+	RunCmd *RunArgs `arg:"subcommand:run" help:"run code on this machine"`
+
+	DeployCmd *DeployArgs `arg:"subcommand:deploy" help:"deploy code into a cluster"`
+
+	// This never runs, it gets preempted in the real main() function.
+	// XXX: Can we do it nicely with the new arg parser? can it ignore all args?
+	EtcdCmd *EtcdArgs `arg:"subcommand:etcd" help:"run standalone etcd"`
+
+	// version is a private handle for our version string.
+	version string `arg:"-"` // ignored from parsing
+
+	// description is a private handle for our description string.
+	description string `arg:"-"` // ignored from parsing
+}
+
+// Version returns the version string. Implementing this signature is part of
+// the API for the cli library.
+func (obj *Args) Version() string {
+	return obj.version
+}
+
+// Description returns a description string. Implementing this signature is part
+// of the API for the cli library.
+func (obj *Args) Description() string {
+	return obj.description
+}
+
+// Run executes the correct subcommand. It errors if there's ever an error. It
+// returns true if we did activate one of the subcommands. It returns false if
+// we did not. This information is used so that the top-level parser can return
+// usage or help information if no subcommand activates.
+func (obj *Args) Run(ctx context.Context, data *cliUtil.Data) (bool, error) {
+	if cmd := obj.RunCmd; cmd != nil {
+		return cmd.Run(ctx, data)
+	}
+
+	if cmd := obj.DeployCmd; cmd != nil {
+		return cmd.Run(ctx, data)
+	}
+
+	// NOTE: we could return true, fmt.Errorf("...") if more than one did
+	return false, nil // nobody activated
+}
+
+// EtcdArgs is the CLI parsing structure and type of the parsed result. This
+// particular one is empty because the `etcd` subcommand is preempted in the
+// real main() function.
+type EtcdArgs struct{}

cli/deploy.go

@@ -0,0 +1,255 @@
+// Mgmt
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
+
+package cli
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"os/signal"
+
+	cliUtil "github.com/purpleidea/mgmt/cli/util"
+	"github.com/purpleidea/mgmt/etcd/client"
+	"github.com/purpleidea/mgmt/etcd/deployer"
+	etcdfs "github.com/purpleidea/mgmt/etcd/fs"
+	"github.com/purpleidea/mgmt/gapi"
+	"github.com/purpleidea/mgmt/lib"
+	"github.com/purpleidea/mgmt/util"
+	"github.com/purpleidea/mgmt/util/errwrap"
+
+	git "github.com/go-git/go-git/v5"
+	"github.com/google/uuid"
+)
+
+// DeployArgs is the CLI parsing structure and type of the parsed result. This
+// particular one contains all the common flags for the `deploy` subcommand
+// which all frontends can use.
+type DeployArgs struct {
+	Seeds []string `arg:"--seeds,env:MGMT_SEEDS" help:"default etc client endpoint"`
+	Noop  bool     `arg:"--noop" help:"globally force all resources into no-op mode"`
+	Sema  int      `arg:"--sema" default:"-1" help:"globally add a semaphore to all resources with this lock count"`
+	NoGit bool     `arg:"--no-git" help:"don't look at git commit id for safe deploys"`
+	Force bool     `arg:"--force" help:"force a new deploy, even if the safety chain would break"`
+
+	DeployEmpty *cliUtil.EmptyArgs `arg:"subcommand:empty" help:"deploy empty payload"`
+	DeployLang  *cliUtil.LangArgs  `arg:"subcommand:lang" help:"deploy lang (mcl) payload"`
+	DeployYaml  *cliUtil.YamlArgs  `arg:"subcommand:yaml" help:"deploy yaml graph payload"`
+}
+
+// Run executes the correct subcommand. It errors if there's ever an error. It
+// returns true if we did activate one of the subcommands. It returns false if
+// we did not. This information is used so that the top-level parser can return
+// usage or help information if no subcommand activates. This particular Run is
+// the run for the main `deploy` subcommand. This always requires a frontend to
+// deploy to the cluster, but if you don't want a graph, you can use the `empty`
+// frontend. The engine backend is agnostic to which frontend is deployed, in
+// fact, you can deploy with multiple different frontends, one after another, on
+// the same engine.
+func (obj *DeployArgs) Run(ctx context.Context, data *cliUtil.Data) (bool, error) {
+	var name string
+	var args interface{}
+	if cmd := obj.DeployEmpty; cmd != nil {
+		name = cliUtil.LookupSubcommand(obj, cmd) // "empty"
+		args = cmd
+	}
+	if cmd := obj.DeployLang; cmd != nil {
+		name = cliUtil.LookupSubcommand(obj, cmd) // "lang"
+		args = cmd
+	}
+	if cmd := obj.DeployYaml; cmd != nil {
+		name = cliUtil.LookupSubcommand(obj, cmd) // "yaml"
+		args = cmd
+	}
+
+	// XXX: workaround https://github.com/alexflint/go-arg/issues/239
+	gapiNames := gapi.Names() // list of registered names
+	if l := len(obj.Seeds); name == "" && l > 1 {
+		elem := obj.Seeds[l-2] // second to last element
+		if util.StrInList(elem, gapiNames) {
+			return false, cliUtil.CliParseError(cliUtil.MissingEquals) // consistent errors
+		}
+	}
+
+	fn, exists := gapi.RegisteredGAPIs[name]
+	if !exists {
+		return false, nil // did not activate
+	}
+	gapiObj := fn()
+
+	program, version := data.Program, data.Version
+	Logf := func(format string, v ...interface{}) {
+		data.Flags.Logf("deploy: "+format, v...)
+	}
+
+	// TODO: consider adding a timeout based on an args.Timeout flag ?
+	ctx, cancel := signal.NotifyContext(ctx, os.Interrupt)
+	defer cancel()
+
+	cliUtil.Hello(program, version, data.Flags) // say hello!
+	defer Logf("goodbye!")
+
+	var hash, pHash string
+	if !obj.NoGit {
+		wd, err := os.Getwd()
+		if err != nil {
+			return false, errwrap.Wrapf(err, "could not get current working directory")
+		}
+		repo, err := git.PlainOpen(wd)
+		if err != nil {
+			return false, errwrap.Wrapf(err, "could not open git repo")
+		}
+
+		head, err := repo.Head()
+		if err != nil {
+			return false, errwrap.Wrapf(err, "could not read git HEAD")
+		}
+
+		hash = head.Hash().String() // current commit id
+		Logf("hash: %s", hash)
+
+		lo := &git.LogOptions{
+			From: head.Hash(),
+		}
+		commits, err := repo.Log(lo)
+		if err != nil {
+			return false, errwrap.Wrapf(err, "could not read git log")
+		}
+		if _, err := commits.Next(); err != nil { // skip over HEAD
+			return false, errwrap.Wrapf(err, "could not read HEAD in git log") // weird!
+		}
+		commit, err := commits.Next()
+		if err == nil { // errors are okay, we might be empty
+			pHash = commit.Hash.String() // previous commit id
+		}
+		Logf("previous deploy hash: %s", pHash)
+		if obj.Force {
+			pHash = "" // don't check this :(
+		}
+		if hash == "" {
+			return false, errwrap.Wrapf(err, "could not get git deploy hash")
+		}
+	}
+
+	uniqueid := uuid.New() // panic's if it can't generate one :P
+
+	etcdClient := client.NewClientFromSeedsNamespace(
+		obj.Seeds, // endpoints
+		lib.NS,
+	)
+	if err := etcdClient.Init(); err != nil {
+		return false, errwrap.Wrapf(err, "client Init failed")
+	}
+	defer func() {
+		err := errwrap.Wrapf(etcdClient.Close(), "client Close failed")
+		if err != nil {
+			// TODO: cause the final exit code to be non-zero
+			Logf("client cleanup error: %+v", err)
+		}
+	}()
+
+	simpleDeploy := &deployer.SimpleDeploy{
+		Client: etcdClient,
+		Debug:  data.Flags.Debug,
+		Logf: func(format string, v ...interface{}) {
+			Logf("deploy: "+format, v...)
+		},
+	}
+	if err := simpleDeploy.Init(); err != nil {
+		return false, errwrap.Wrapf(err, "deploy Init failed")
+	}
+	defer func() {
+		err := errwrap.Wrapf(simpleDeploy.Close(), "deploy Close failed")
+		if err != nil {
+			// TODO: cause the final exit code to be non-zero
+			Logf("deploy cleanup error: %+v", err)
+		}
+	}()
+
+	// get max id (from all the previous deploys)
+	max, err := simpleDeploy.GetMaxDeployID(ctx)
+	if err != nil {
+		return false, errwrap.Wrapf(err, "error getting max deploy id")
+	}
+	// find the latest id
+	var id = max + 1 // next id
+	Logf("previous max deploy id: %d", max)
+
+	etcdFs := &etcdfs.Fs{
+		Client: etcdClient,
+		// TODO: using a uuid is meant as a temporary measure, i hate them
+		Metadata:   lib.MetadataPrefix + fmt.Sprintf("/deploy/%d-%s", id, uniqueid),
+		DataPrefix: lib.StoragePrefix,
+
+		Debug: data.Flags.Debug,
+		Logf: func(format string, v ...interface{}) {
+			Logf("fs: "+format, v...)
+		},
+	}
+
+	info := &gapi.Info{
+		Args: args,
+		Flags: &gapi.Flags{
+			Noop: obj.Noop,
+			Sema: obj.Sema,
+			//Update: obj.Update,
+		},
+
+		Fs:    etcdFs,
+		Debug: data.Flags.Debug,
+		Logf: func(format string, v ...interface{}) {
+			// TODO: is this a sane prefix to use here?
+			data.Flags.Logf("cli: "+format, v...)
+		},
+	}
+
+	deploy, err := gapiObj.Cli(info)
+	if err != nil {
+		return false, cliUtil.CliParseError(err) // consistent errors
+	}
+	if deploy == nil { // not used
+		return false, fmt.Errorf("not enough information specified")
+	}
+
+	// redundant
+	deploy.Noop = obj.Noop
+	deploy.Sema = obj.Sema
+
+	str, err := deploy.ToB64()
+	if err != nil {
+		return false, errwrap.Wrapf(err, "encoding error")
+	}
+
+	// this nominally checks the previous git hash matches our expectation
+	if err := simpleDeploy.AddDeploy(ctx, id, hash, pHash, &str); err != nil {
+		return false, errwrap.Wrapf(err, "could not create deploy id `%d`", id)
+	}
+	Logf("success, id: %d", id)
+	return true, nil
+}

cli/run.go

@@ -0,0 +1,233 @@
+// Mgmt
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
+
+package cli
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"os/signal"
+	"sync"
+	"syscall"
+
+	cliUtil "github.com/purpleidea/mgmt/cli/util"
+	"github.com/purpleidea/mgmt/gapi"
+	"github.com/purpleidea/mgmt/lib"
+	"github.com/purpleidea/mgmt/util"
+	"github.com/purpleidea/mgmt/util/errwrap"
+
+	"github.com/spf13/afero"
+)
+
+// RunArgs is the CLI parsing structure and type of the parsed result. This
+// particular one contains all the common flags for the `run` subcommand which
+// all frontends can use.
+type RunArgs struct {
+	lib.Config // embedded config (can't be a pointer) https://github.com/alexflint/go-arg/issues/240
+
+	RunEmpty *cliUtil.EmptyArgs `arg:"subcommand:empty" help:"run empty payload"`
+	RunLang  *cliUtil.LangArgs  `arg:"subcommand:lang" help:"run lang (mcl) payload"`
+	RunYaml  *cliUtil.YamlArgs  `arg:"subcommand:yaml" help:"run yaml graph payload"`
+}
+
+// Run executes the correct subcommand. It errors if there's ever an error. It
+// returns true if we did activate one of the subcommands. It returns false if
+// we did not. This information is used so that the top-level parser can return
+// usage or help information if no subcommand activates. This particular Run is
+// the run for the main `run` subcommand. This always requires a frontend to
+// start the engine, but if you don't want a graph, you can use the `empty`
+// frontend. The engine backend is agnostic to which frontend is running, in
+// fact, you can deploy with multiple different frontends, one after another, on
+// the same engine.
+func (obj *RunArgs) Run(ctx context.Context, data *cliUtil.Data) (bool, error) {
+	var name string
+	var args interface{}
+	if cmd := obj.RunEmpty; cmd != nil {
+		name = cliUtil.LookupSubcommand(obj, cmd) // "empty"
+		args = cmd
+	}
+	if cmd := obj.RunLang; cmd != nil {
+		name = cliUtil.LookupSubcommand(obj, cmd) // "lang"
+		args = cmd
+	}
+	if cmd := obj.RunYaml; cmd != nil {
+		name = cliUtil.LookupSubcommand(obj, cmd) // "yaml"
+		args = cmd
+	}
+
+	// XXX: workaround https://github.com/alexflint/go-arg/issues/239
+	lists := [][]string{
+		obj.Seeds,
+		obj.ClientURLs,
+		obj.ServerURLs,
+		obj.AdvertiseClientURLs,
+		obj.AdvertiseServerURLs,
+	}
+	gapiNames := gapi.Names() // list of registered names
+	for _, list := range lists {
+		if l := len(list); name == "" && l > 1 {
+			elem := list[l-2] // second to last element
+			if util.StrInList(elem, gapiNames) {
+				return false, cliUtil.CliParseError(cliUtil.MissingEquals) // consistent errors
+			}
+		}
+	}
+
+	fn, exists := gapi.RegisteredGAPIs[name]
+	if !exists {
+		return false, nil // did not activate
+	}
+	gapiObj := fn()
+
+	main := &lib.Main{}
+	main.Config = &obj.Config // pass in all the parsed data
+
+	main.Program, main.Version = data.Program, data.Version
+	main.Debug, main.Logf = data.Flags.Debug, data.Flags.Logf // no prefix
+	Logf := func(format string, v ...interface{}) {
+		data.Flags.Logf("main: "+format, v...)
+	}
+
+	cliUtil.Hello(main.Program, main.Version, data.Flags) // say hello!
+	defer Logf("goodbye!")
+
+	// create a memory backed temporary filesystem for storing runtime data
+	mmFs := afero.NewMemMapFs()
+	afs := &afero.Afero{Fs: mmFs} // wrap so that we're implementing ioutil
+	standaloneFs := &util.AferoFs{Afero: afs}
+	main.DeployFs = standaloneFs
+
+	info := &gapi.Info{
+		Args: args,
+		Flags: &gapi.Flags{
+			Hostname: obj.Hostname,
+			Noop:     obj.Noop,
+			Sema:     obj.Sema,
+			//Update: obj.Update,
+		},
+
+		Fs:    standaloneFs,
+		Debug: data.Flags.Debug,
+		Logf: func(format string, v ...interface{}) {
+			data.Flags.Logf("cli: "+format, v...)
+		},
+	}
+
+	deploy, err := gapiObj.Cli(info)
+	if err != nil {
+		return false, cliUtil.CliParseError(err) // consistent errors
+	}
+
+	if cmd := obj.RunLang; cmd != nil && cmd.OnlyUnify && deploy == nil {
+		return true, nil // we end early
+	}
+	if cmd := obj.RunLang; cmd != nil && cmd.OnlyDownload && deploy == nil {
+		return true, nil // we end early
+	}
+	main.Deploy = deploy
+	if main.Deploy == nil {
+		// nobody activated, but we'll still watch the etcd deploy chan,
+		// and if there is deployed code that's ready to run, we'll run!
+		data.Flags.Logf("main: no frontend selected (no GAPI activated)")
+	}
+
+	if err := main.Validate(); err != nil {
+		return false, err
+	}
+
+	if err := main.Init(); err != nil {
+		return false, err
+	}
+
+	// install the exit signal handler
+	wg := &sync.WaitGroup{}
+	defer wg.Wait()
+	exit := make(chan struct{})
+	defer close(exit)
+	wg.Add(1)
+	go func() {
+		defer wg.Done()
+		// must have buffer for max number of signals
+		signals := make(chan os.Signal, 3+1) // 3 * ^C + 1 * SIGTERM
+		signal.Notify(signals, os.Interrupt) // catch ^C
+		//signal.Notify(signals, os.Kill) // catch signals
+		signal.Notify(signals, syscall.SIGTERM)
+		var count uint8
+		for {
+			select {
+			case sig := <-signals: // any signal will do
+				if sig != os.Interrupt {
+					data.Flags.Logf("interrupted by signal")
+					main.Interrupt(fmt.Errorf("killed by %v", sig))
+					return
+				}
+
+				switch count {
+				case 0:
+					data.Flags.Logf("interrupted by ^C")
+					main.Exit(nil)
+				case 1:
+					data.Flags.Logf("interrupted by ^C (fast pause)")
+					main.FastExit(nil)
+				case 2:
+					data.Flags.Logf("interrupted by ^C (hard interrupt)")
+					main.Interrupt(nil)
+				}
+				count++
+
+			case <-exit:
+				return
+			}
+		}
+	}()
+
+	reterr := main.Run()
+	if reterr != nil {
+		// log the error message returned
+		if data.Flags.Debug {
+			data.Flags.Logf("main: %+v", reterr)
+		}
+	}
+
+	if err := main.Close(); err != nil {
+		if data.Flags.Debug {
+			data.Flags.Logf("main: Close: %+v", err)
+		}
+		if reterr == nil {
+			return false, err
+		}
+		reterr = errwrap.Append(reterr, err)
+	}
+
+	if reterr != nil {
+		return false, reterr
+	}
+	return true, nil
+}

cli/util/args.go

@@ -0,0 +1,102 @@
+// Mgmt
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
+
+package util
+
+import (
+	"reflect"
+	"strings"
+)
+
+// LookupSubcommand returns the name of the subcommand in the obj, of a struct.
+// This is useful for determining the name of the subcommand that was activated.
+// It returns an empty string if a specific name was not found.
+func LookupSubcommand(obj interface{}, st interface{}) string {
+	val := reflect.ValueOf(obj)
+	if val.Kind() == reflect.Ptr { // max one de-referencing
+		val = val.Elem()
+	}
+
+	v := reflect.ValueOf(st) // value of the struct
+	typ := val.Type()
+	for i := 0; i < typ.NumField(); i++ {
+		f := val.Field(i) // value of the field
+		if f.Interface() != v.Interface() {
+			continue
+		}
+
+		field := typ.Field(i)
+		alias, ok := field.Tag.Lookup("arg")
+		if !ok {
+			continue
+		}
+
+		// XXX: `arg` needs a split by comma first or fancier parsing
+		prefix := "subcommand"
+		split := strings.Split(alias, ":")
+		if len(split) != 2 || split[0] != prefix {
+			continue
+		}
+
+		return split[1] // found
+	}
+	return "" // not found
+}
+
+// EmptyArgs is the empty CLI parsing structure and type of the parsed result.
+type EmptyArgs struct{}
+
+// LangArgs is the lang CLI parsing structure and type of the parsed result.
+type LangArgs struct {
+	// Input is the input mcl code or file path or any input specification.
+	Input string `arg:"positional,required"`
+
+	// TODO: removed (temporarily?)
+	//Stdin bool `arg:"--stdin" help:"use passthrough stdin"`
+
+	Download     bool `arg:"--download" help:"download any missing imports"`
+	OnlyDownload bool `arg:"--only-download" help:"stop after downloading any missing imports"`
+	Update       bool `arg:"--update" help:"update all dependencies to the latest versions"`
+
+	OnlyUnify bool `arg:"--only-unify" help:"stop after type unification"`
+	SkipUnify bool `arg:"--skip-unify" help:"skip type unification"`
+
+	Depth int `arg:"--depth" default:"-1" help:"max recursion depth limit (-1 is unlimited)"`
+
+	// The default of 0 means any error is a failure by default.
+	Retry int `arg:"--depth" help:"max number of retries (-1 is unlimited)"`
+
+	ModulePath string `arg:"--module-path,env:MGMT_MODULE_PATH" help:"choose the modules path (absolute)"`
+}
+
+// YamlArgs is the yaml CLI parsing structure and type of the parsed result.
+type YamlArgs struct {
+	// Input is the input yaml code or file path or any input specification.
+	Input string `arg:"positional,required"`
+}

cli/util/hello.go

@@ -0,0 +1,47 @@
+// Mgmt
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
+
+package util
+
+import (
+	"fmt"
+	"time"
+)
+
+// Hello is a simple helper function to print a hello message and time.
+func Hello(program, version string, flags Flags) {
+	var start = time.Now().UnixNano()
+	if program == "" {
+		program = "<unknown>"
+	}
+	fmt.Println(fmt.Sprintf("This is: %s, version: %s", program, version))
+	fmt.Println("Copyright (C) 2013-2024+ James Shubin and the project contributors")
+	fmt.Println("Written by James Shubin <james@shubin.ca> and the project contributors")
+	flags.Logf("main: start: %v", start)
+}

cli/util/util.go

@@ -0,0 +1,98 @@
+// Mgmt
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
+
+// Package util has some CLI related utility code.
+package util
+
+import (
+	"log"
+	"os"
+	"strings"
+
+	"github.com/purpleidea/mgmt/util/errwrap"
+)
+
+// Error is a constant error type that implements error.
+type Error string
+
+// Error fulfills the error interface of this type.
+func (e Error) Error() string { return string(e) }
+
+const (
+	// MissingEquals means we probably hit the parsing bug.
+	// XXX: see: https://github.com/alexflint/go-arg/issues/239
+	MissingEquals = Error("missing equals sign for list element")
+)
+
+// CliParseError returns a consistent error if we have a CLI parsing issue.
+func CliParseError(err error) error {
+	return errwrap.Wrapf(err, "cli parse error")
+}
+
+// Flags are some constant flags which are used throughout the program.
+type Flags struct {
+	Debug bool // add additional log messages
+	Logf  func(format string, v ...interface{})
+}
+
+// Data is a struct of values that we usually pass to the main CLI function.
+type Data struct {
+	Program string
+	Version string
+	Copying string
+	Tagline string
+	Flags   Flags
+	Args    []string // os.Args usually
+}
+
+// SafeProgram returns the correct program string when given a buggy variant.
+func SafeProgram(program string) string {
+	// FIXME: in sub commands, the cli package appends a space and the sub
+	// command name at the end. hack around this by only using the first bit
+	// see: https://github.com/urfave/cli/issues/783 for more details...
+	split := strings.Split(program, " ")
+	program = split[0]
+	//if program == "" {
+	//	program = "<unknown>"
+	//}
+	return program
+}
+
+// LogSetup changes some of the core logger package settings.
+func LogSetup(debug bool) {
+	// TODO: Move these log package initialization steps to the top main.go?
+	logFlags := log.LstdFlags
+	if debug {
+		logFlags = logFlags + log.Lshortfile
+	}
+	logFlags = logFlags - log.Ldate // remove the date for now
+	log.SetFlags(logFlags)
+
+	log.SetOutput(os.Stderr)
+}

converger/converger.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,7 +13,19 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 // Package converger is a facility for reporting the converged state.
 package converger
@@ -25,139 +37,251 @@ import (
 	"time"
 
 	"github.com/purpleidea/mgmt/util"
-
-	multierr "github.com/hashicorp/go-multierror"
+	"github.com/purpleidea/mgmt/util/errwrap"
 )
 
-// TODO: we could make a new function that masks out the state of certain
-// UID's, but at the moment the new Timer code has obsoleted the need...
+// New builds a new converger coordinator.
+func New(timeout int) *Coordinator {
+	return &Coordinator{
+		timeout: timeout,
 
-// Converger is the general interface for implementing a convergence watcher.
-type Converger interface { // TODO: need a better name
-	Register() UID
-	IsConverged(UID) bool         // is the UID converged ?
-	SetConverged(UID, bool) error // set the converged state of the UID
-	Unregister(UID)
-	Start()
-	Pause()
-	Loop(bool)
-	ConvergedTimer(UID) <-chan time.Time
-	Status() map[uint64]bool
-	Timeout() int                              // returns the timeout that this was created with
-	AddStateFn(string, func(bool) error) error // adds a stateFn with a name
-	RemoveStateFn(string) error                // remove a stateFn with a given name
-}
+		mutex: &sync.RWMutex{},
 
-// UID is the interface resources can use to notify with if converged. You'll
-// need to use part of the Converger interface to Register initially too.
-type UID interface {
-	ID() uint64   // get Id
-	Name() string // get a friendly name
-	SetName(string)
-	IsValid() bool // has Id been initialized ?
-	InvalidateID() // set Id to nil
-	IsConverged() bool
-	SetConverged(bool) error
-	Unregister()
-	ConvergedTimer() <-chan time.Time
-	StartTimer() (func() error, error) // cancellable is the same as StopTimer()
-	ResetTimer() error                 // resets counter to zero
-	StopTimer() error
-}
+		//lastid: 0,
+		status: make(map[*UID]struct{}),
 
-// converger is an implementation of the Converger interface.
-type converger struct {
-	timeout   int           // must be zero (instant) or greater seconds to run
-	converged bool          // did we converge (state changes of this run Fn)
-	channel   chan struct{} // signal here to run an isConverged check
-	control   chan bool     // control channel for start/pause
-	mutex     *sync.RWMutex // used for controlling access to status and lastid
-	lastid    uint64
-	status    map[uint64]bool
-	stateFns  map[string]func(bool) error // run on converged state changes with state bool
-	smutex    *sync.RWMutex               // used for controlling access to stateFns
-}
+		//converged: false, // initial state
 
-// cuid is an implementation of the UID interface.
-type cuid struct {
-	converger Converger
-	id        uint64
-	name      string // user defined, friendly name
-	mutex     *sync.Mutex
-	timer     chan struct{}
-	running   bool // is the above timer running?
-	wg        *sync.WaitGroup
-}
+		pokeChan: make(chan struct{}, 1), // must be buffered
+
+		readyChan: make(chan struct{}), // ready signal
+
+		//paused: false, // starts off as started
+		pauseSignal: make(chan struct{}),
+		//resumeSignal: make(chan struct{}), // happens on pause
+		//pausedAck: util.NewEasyAck(), // happens on pause
 
-// NewConverger builds a new converger struct.
-func NewConverger(timeout int) Converger {
-	return &converger{
-		timeout:  timeout,
-		channel:  make(chan struct{}),
-		control:  make(chan bool),
-		mutex:    &sync.RWMutex{},
-		lastid:   0,
-		status:   make(map[uint64]bool),
 		stateFns: make(map[string]func(bool) error),
 		smutex:   &sync.RWMutex{},
-	}
-}
 
-// Register assigns a UID to the caller.
-func (obj *converger) Register() UID {
-	obj.mutex.Lock()
-	defer obj.mutex.Unlock()
-	obj.lastid++
-	obj.status[obj.lastid] = false // initialize as not converged
-	return &cuid{
-		converger: obj,
-		id:        obj.lastid,
-		name:      fmt.Sprintf("%d", obj.lastid), // some default
-		mutex:     &sync.Mutex{},
-		timer:     nil,
-		running:   false,
+		closeChan: make(chan struct{}),
 		wg:        &sync.WaitGroup{},
 	}
 }
 
-// IsConverged gets the converged status of a uid.
-func (obj *converger) IsConverged(uid UID) bool {
-	if !uid.IsValid() {
-		panic(fmt.Sprintf("the ID of UID(%s) is nil", uid.Name()))
-	}
-	obj.mutex.RLock()
-	isConverged, found := obj.status[uid.ID()] // lookup
-	obj.mutex.RUnlock()
-	if !found {
-		panic("the ID of UID is unregistered")
-	}
-	return isConverged
+// Coordinator is the central converger engine.
+type Coordinator struct {
+	// timeout must be zero (instant) or greater seconds to run. If it's -1
+	// then this is disabled, and we never run stateFns.
+	timeout int
+
+	// mutex is used for controlling access to status and lastid.
+	mutex *sync.RWMutex
+
+	// lastid contains the last uid we used for registration.
+	//lastid    uint64
+	// status contains a reference to each active UID.
+	status map[*UID]struct{}
+
+	// converged stores the last convergence state. When this changes, we
+	// run the stateFns.
+	converged bool
+
+	// pokeChan receives a message every time we might need to re-calculate.
+	pokeChan chan struct{}
+
+	// readyChan closes to notify any interested parties that the main loop
+	// is running.
+	readyChan chan struct{}
+
+	// paused represents if this coordinator is paused or not.
+	paused bool
+	// pauseSignal closes to request a pause of this coordinator.
+	pauseSignal chan struct{}
+	// resumeSignal closes to request a resume of this coordinator.
+	resumeSignal chan struct{}
+	// pausedAck is used to send an ack message saying that we've paused.
+	pausedAck *util.EasyAck
+
+	// stateFns run on converged state changes.
+	stateFns map[string]func(bool) error
+	// smutex is used for controlling access to the stateFns map.
+	smutex *sync.RWMutex
+
+	// closeChan closes when we've been requested to shutdown.
+	closeChan chan struct{}
+	// wg waits for everything to finish.
+	wg *sync.WaitGroup
 }
 
-// SetConverged updates the converger with the converged state of the UID.
-func (obj *converger) SetConverged(uid UID, isConverged bool) error {
-	if !uid.IsValid() {
-		return fmt.Errorf("the ID of UID(%s) is nil", uid.Name())
-	}
+// Register creates a new UID which can be used to report converged state. You
+// must Unregister each UID before Shutdown will be able to finish running.
+func (obj *Coordinator) Register() *UID {
+	obj.wg.Add(1) // additional tracking for each UID
 	obj.mutex.Lock()
-	if _, found := obj.status[uid.ID()]; !found {
-		panic("the ID of UID is unregistered")
+	defer obj.mutex.Unlock()
+	//obj.lastid++
+	uid := &UID{
+		timeout: obj.timeout, // copy the timeout here
+		//id: obj.lastid,
+		//name: fmt.Sprintf("%d", obj.lastid), // some default
+
+		poke: obj.poke,
+
+		// timer
+		mutex:   &sync.Mutex{},
+		timer:   nil,
+		running: false,
+		wg:      &sync.WaitGroup{},
 	}
-	obj.status[uid.ID()] = isConverged // set
-	obj.mutex.Unlock()                 // unlock *before* poke or deadlock!
-	if isConverged != obj.converged {  // only poke if it would be helpful
-		// run in a go routine so that we never block... just queue up!
-		// this allows us to send events, even if we haven't started...
-		go func() { obj.channel <- struct{}{} }()
+	uid.unregister = func() { obj.Unregister(uid) } // add unregister func
+	obj.status[uid] = struct{}{}                    // TODO: add converged state here?
+	return uid
+}
+
+// Unregister removes the UID from the converger coordinator. If you supply an
+// invalid or unregistered uid to this function, it will panic. An unregistered
+// UID is no longer part of the convergence checking.
+func (obj *Coordinator) Unregister(uid *UID) {
+	defer obj.wg.Done() // additional tracking for each UID
+	obj.mutex.Lock()
+	defer obj.mutex.Unlock()
+
+	if _, exists := obj.status[uid]; !exists {
+		panic("uid is not registered")
 	}
+	uid.StopTimer() // ignore any errors
+	delete(obj.status, uid)
+}
+
+// Run starts the main loop for the converger coordinator. It is commonly run
+// from a go routine. It blocks until the Shutdown method is run to close it.
+// NOTE: when we have very short timeouts, if we start before all the resources
+// have joined the map, then it might appear as if we converged before we did!
+func (obj *Coordinator) Run(startPaused bool) {
+	obj.wg.Add(1)
+	wg := &sync.WaitGroup{} // needed for the startPaused
+	defer wg.Wait()         // don't leave any leftover go routines running
+	if startPaused {
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			obj.Pause() // ignore any errors
+			close(obj.readyChan)
+		}()
+	} else {
+		close(obj.readyChan) // we must wait till the wg.Add(1) has happened...
+	}
+	defer obj.wg.Done()
+	for {
+		// pause if one was requested...
+		select {
+		case <-obj.pauseSignal: // channel closes
+			obj.pausedAck.Ack() // send ack
+			// we are paused now, and waiting for resume or exit...
+			select {
+			case <-obj.resumeSignal: // channel closes
+				// resumed!
+
+			case <-obj.closeChan: // we can always escape
+				return
+			}
+
+		case _, ok := <-obj.pokeChan: // we got an event (re-calculate)
+			if !ok {
+				return
+			}
+
+			if err := obj.test(); err != nil {
+				// FIXME: what to do on error ?
+			}
+
+		case <-obj.closeChan: // we can always escape
+			return
+		}
+	}
+}
+
+// Ready blocks until the Run loop has started up. This is useful so that we
+// don't run Shutdown before we've even started up properly.
+func (obj *Coordinator) Ready() {
+	select {
+	case <-obj.readyChan:
+	}
+}
+
+// Shutdown sends a signal to the Run loop that it should exit. This blocks
+// until it does.
+func (obj *Coordinator) Shutdown() {
+	close(obj.closeChan)
+	obj.wg.Wait()
+	close(obj.pokeChan) // free memory?
+}
+
+// Pause pauses the coordinator. It should not be called on an already paused
+// coordinator. It will block until the coordinator pauses with an
+// acknowledgment, or until an exit is requested. If the latter happens it will
+// error. It is NOT thread-safe with the Resume() method so only call either one
+// at a time.
+func (obj *Coordinator) Pause() error {
+	if obj.paused {
+		return fmt.Errorf("already paused")
+	}
+
+	obj.pausedAck = util.NewEasyAck()
+	obj.resumeSignal = make(chan struct{}) // build the resume signal
+	close(obj.pauseSignal)
+
+	// wait for ack (or exit signal)
+	select {
+	case <-obj.pausedAck.Wait(): // we got it!
+		// we're paused
+	case <-obj.closeChan:
+		return fmt.Errorf("closing")
+	}
+	obj.paused = true
+
 	return nil
 }
 
-// isConverged returns true if *every* registered uid has converged.
-func (obj *converger) isConverged() bool {
-	obj.mutex.RLock() // take a read lock
-	defer obj.mutex.RUnlock()
-	for _, v := range obj.status {
+// Resume unpauses the coordinator. It can be safely called on a brand-new
+// coordinator that has just started running without incident. It is NOT
+// thread-safe with the Pause() method, so only call either one at a time.
+func (obj *Coordinator) Resume() {
+	// TODO: do we need a mutex around Resume?
+	if !obj.paused { // no need to unpause brand-new resources
+		return
+	}
+
+	obj.pauseSignal = make(chan struct{}) // rebuild for next pause
+	close(obj.resumeSignal)
+	obj.poke() // unblock and notice the resume if necessary
+
+	obj.paused = false
+
+	// no need to wait for it to resume
+	//return // implied
+}
+
+// poke sends a message to the coordinator telling it that it should re-evaluate
+// whether we're converged or not. This does not block. Do not run this in a
+// goroutine. It must not be called after Shutdown has been called.
+func (obj *Coordinator) poke() {
+	// redundant
+	//if len(obj.pokeChan) > 0 {
+	//	return
+	//}
+
+	select {
+	case obj.pokeChan <- struct{}{}:
+	default: // if chan is now full because more than one poke happened...
+	}
+}
+
+// IsConverged returns true if *every* registered uid has converged. If there
+// are no registered UID's, then this will return true.
+func (obj *Coordinator) IsConverged() bool {
+	for _, v := range obj.Status() {
 		if !v { // everyone must be converged for this to be true
 			return false
 		}
@@ -165,145 +289,40 @@ func (obj *converger) isConverged() bool {
 	return true
 }
 
-// Unregister dissociates the ConvergedUID from the converged checking.
-func (obj *converger) Unregister(uid UID) {
-	if !uid.IsValid() {
-		panic(fmt.Sprintf("the ID of UID(%s) is nil", uid.Name()))
+// test evaluates whether we're converged or not and runs the state change. It
+// is NOT thread-safe.
+func (obj *Coordinator) test() error {
+	// TODO: add these checks elsewhere to prevent anything from running?
+	if obj.timeout < 0 {
+		return nil // nothing to do (only run if timeout is valid)
 	}
-	obj.mutex.Lock()
-	uid.StopTimer() // ignore any errors
-	delete(obj.status, uid.ID())
-	obj.mutex.Unlock()
-	uid.InvalidateID()
-}
 
-// Start causes a Converger object to start or resume running.
-func (obj *converger) Start() {
-	obj.control <- true
-}
+	converged := obj.IsConverged()
+	defer func() {
+		obj.converged = converged // set this only at the end...
+	}()
 
-// Pause causes a Converger object to stop running temporarily.
-func (obj *converger) Pause() { // FIXME: add a sync ACK on pause before return
-	obj.control <- false
-}
-
-// Loop is the main loop for a Converger object. It usually runs in a goroutine.
-// TODO: we could eventually have each resource tell us as soon as it converges,
-// and then keep track of the time delays here, to avoid callers needing select.
-// NOTE: when we have very short timeouts, if we start before all the resources
-// have joined the map, then it might appear as if we converged before we did!
-func (obj *converger) Loop(startPaused bool) {
-	if obj.control == nil {
-		panic("converger not initialized correctly")
-	}
-	if startPaused { // start paused without racing
-		select {
-		case e := <-obj.control:
-			if !e {
-				panic("converger expected true")
-			}
+	if !converged {
+		if !obj.converged { // were we previously also not converged?
+			return nil // nothing to do
 		}
-	}
-	for {
-		select {
-		case e := <-obj.control: // expecting "false" which means pause!
-			if e {
-				panic("converger expected false")
-			}
-			// now i'm paused...
-			select {
-			case e := <-obj.control:
-				if !e {
-					panic("converger expected true")
-				}
-				// restart
-				// kick once to refresh the check...
-				go func() { obj.channel <- struct{}{} }()
-				continue
-			}
 
-		case <-obj.channel:
-			if !obj.isConverged() {
-				if obj.converged { // we're doing a state change
-					// call the arbitrary functions (takes a read lock!)
-					if err := obj.runStateFns(false); err != nil {
-						// FIXME: what to do on error ?
-					}
-				}
-				obj.converged = false
-				continue
-			}
-
-			// we have converged!
-			if obj.timeout >= 0 { // only run if timeout is valid
-				if !obj.converged { // we're doing a state change
-					// call the arbitrary functions (takes a read lock!)
-					if err := obj.runStateFns(true); err != nil {
-						// FIXME: what to do on error ?
-					}
-				}
-			}
-			obj.converged = true
-			// loop and wait again...
-		}
+		// we're doing a state change
+		// call the arbitrary functions (takes a read lock!)
+		return obj.runStateFns(false)
 	}
+
+	// we have converged!
+	if obj.converged { // were we previously also converged?
+		return nil // nothing to do
+	}
+
+	// call the arbitrary functions (takes a read lock!)
+	return obj.runStateFns(true)
 }
 
-// ConvergedTimer adds a timeout to a select call and blocks until then.
-// TODO: this means we could eventually have per resource converged timeouts
-func (obj *converger) ConvergedTimer(uid UID) <-chan time.Time {
-	// be clever: if i'm already converged, this timeout should block which
-	// avoids unnecessary new signals being sent! this avoids fast loops if
-	// we have a low timeout, or in particular a timeout == 0
-	if uid.IsConverged() {
-		// blocks the case statement in select forever!
-		return util.TimeAfterOrBlock(-1)
-	}
-	return util.TimeAfterOrBlock(obj.timeout)
-}
-
-// Status returns a map of the converged status of each UID.
-func (obj *converger) Status() map[uint64]bool {
-	status := make(map[uint64]bool)
-	obj.mutex.RLock() // take a read lock
-	defer obj.mutex.RUnlock()
-	for k, v := range obj.status { // make a copy to avoid the mutex
-		status[k] = v
-	}
-	return status
-}
-
-// Timeout returns the timeout in seconds that converger was created with. This
-// is useful to avoid passing in the timeout value separately when you're
-// already passing in the Converger struct.
-func (obj *converger) Timeout() int {
-	return obj.timeout
-}
-
-// AddStateFn adds a state function to be run on change of converged state.
-func (obj *converger) AddStateFn(name string, stateFn func(bool) error) error {
-	obj.smutex.Lock()
-	defer obj.smutex.Unlock()
-	if _, exists := obj.stateFns[name]; exists {
-		return fmt.Errorf("a stateFn with that name already exists")
-	}
-	obj.stateFns[name] = stateFn
-	return nil
-}
-
-// RemoveStateFn adds a state function to be run on change of converged state.
-func (obj *converger) RemoveStateFn(name string) error {
-	obj.smutex.Lock()
-	defer obj.smutex.Unlock()
-	if _, exists := obj.stateFns[name]; !exists {
-		return fmt.Errorf("a stateFn with that name doesn't exist")
-	}
-	delete(obj.stateFns, name)
-	return nil
-}
-
-// runStateFns runs the listed of stored state functions.
-func (obj *converger) runStateFns(converged bool) error {
+// runStateFns runs the list of stored state functions.
+func (obj *Coordinator) runStateFns(converged bool) error {
 	obj.smutex.RLock()
 	defer obj.smutex.RUnlock()
 	var keys []string
@@ -315,77 +334,125 @@ func (obj *converger) runStateFns(converged bool) error {
 	for _, name := range keys { // run in deterministic order
 		fn := obj.stateFns[name]
 		// call an arbitrary function
-		if e := fn(converged); e != nil {
-			err = multierr.Append(err, e) // list of errors
-		}
+		e := fn(converged)
+		err = errwrap.Append(err, e) // list of errors
 	}
 	return err
 }
 
-// ID returns the unique id of this UID object.
-func (obj *cuid) ID() uint64 {
-	return obj.id
+// AddStateFn adds a state function to be run on change of converged state.
+func (obj *Coordinator) AddStateFn(name string, stateFn func(bool) error) error {
+	obj.smutex.Lock()
+	defer obj.smutex.Unlock()
+	if _, exists := obj.stateFns[name]; exists {
+		return fmt.Errorf("a stateFn with that name already exists")
+	}
+	obj.stateFns[name] = stateFn
+	return nil
 }
 
-// Name returns a user defined name for the specific cuid.
-func (obj *cuid) Name() string {
-	return obj.name
+// RemoveStateFn removes a state function from running on change of converged
+// state.
+func (obj *Coordinator) RemoveStateFn(name string) error {
+	obj.smutex.Lock()
+	defer obj.smutex.Unlock()
+	if _, exists := obj.stateFns[name]; !exists {
+		return fmt.Errorf("a stateFn with that name doesn't exist")
+	}
+	delete(obj.stateFns, name)
+	return nil
 }
 
-// SetName sets a user defined name for the specific cuid.
-func (obj *cuid) SetName(name string) {
-	obj.name = name
+// Status returns a map of the converged status of each UID.
+func (obj *Coordinator) Status() map[*UID]bool {
+	status := make(map[*UID]bool)
+	obj.mutex.RLock() // take a read lock
+	defer obj.mutex.RUnlock()
+	for k := range obj.status {
+		status[k] = k.IsConverged()
+	}
+	return status
 }
 
-// IsValid tells us if the id is valid or has already been destroyed.
-func (obj *cuid) IsValid() bool {
-	return obj.id != 0 // an id of 0 is invalid
+// Timeout returns the timeout in seconds that converger was created with. This
+// is useful to avoid passing in the timeout value separately when you're
+// already passing in the Coordinator struct.
+func (obj *Coordinator) Timeout() int {
+	return obj.timeout
 }
 
-// InvalidateID marks the id as no longer valid.
-func (obj *cuid) InvalidateID() {
-	obj.id = 0 // an id of 0 is invalid
+// UID represents one of the probes for the converger coordinator. It is created
+// by calling the Register method of the Coordinator struct. It should be freed
+// after use with Unregister.
+type UID struct {
+	// timeout is a copy of the main timeout. It could eventually be used
+	// for per-UID timeouts too.
+	timeout int
+	// isConverged stores the convergence state of this particular UID.
+	isConverged bool
+
+	// poke stores a reference to the main poke function.
+	poke func()
+	// unregister stores a reference to the unregister function.
+	unregister func()
+
+	// timer
+	mutex   *sync.Mutex
+	timer   chan struct{}
+	running bool // is the timer running?
+	wg      *sync.WaitGroup
 }
 
-// IsConverged is a helper function to the regular IsConverged method.
-func (obj *cuid) IsConverged() bool {
-	return obj.converger.IsConverged(obj)
+// Unregister removes this UID from the converger coordinator. An unregistered
+// UID is no longer part of the convergence checking.
+func (obj *UID) Unregister() {
+	obj.unregister()
 }
 
-// SetConverged is a helper function to the regular SetConverged notification.
-func (obj *cuid) SetConverged(isConverged bool) error {
-	return obj.converger.SetConverged(obj, isConverged)
+// IsConverged reports whether this UID is converged or not.
+func (obj *UID) IsConverged() bool {
+	return obj.isConverged
 }
 
-// Unregister is a helper function to unregister myself.
-func (obj *cuid) Unregister() {
-	obj.converger.Unregister(obj)
+// SetConverged sets the convergence state of this UID. This is used by the
+// running timer if one is started. The timer will overwrite any value set by
+// this method.
+func (obj *UID) SetConverged(isConverged bool) {
+	obj.isConverged = isConverged
+	obj.poke() // notify of change
 }
 
-// ConvergedTimer is a helper around the regular ConvergedTimer method.
-func (obj *cuid) ConvergedTimer() <-chan time.Time {
-	return obj.converger.ConvergedTimer(obj)
+// ConvergedTimer adds a timeout to a select call and blocks until then.
+// TODO: this means we could eventually have per resource converged timeouts
+func (obj *UID) ConvergedTimer() <-chan time.Time {
+	// be clever: if i'm already converged, this timeout should block which
+	// avoids unnecessary new signals being sent! this avoids fast loops if
+	// we have a low timeout, or in particular a timeout == 0
+	if obj.IsConverged() {
+		// blocks the case statement in select forever!
+		return util.TimeAfterOrBlock(-1)
+	}
+	return util.TimeAfterOrBlock(int(obj.timeout))
 }
 
-// StartTimer runs an invisible timer that automatically converges on timeout.
-func (obj *cuid) StartTimer() (func() error, error) {
+// StartTimer runs a timer that sets us as converged on timeout. It also returns
+// a handle to the StopTimer function which should be run before exit.
+func (obj *UID) StartTimer() (func() error, error) {
 	obj.mutex.Lock()
-	if !obj.running {
-		obj.timer = make(chan struct{})
-		obj.running = true
-	} else {
-		obj.mutex.Unlock()
+	defer obj.mutex.Unlock()
+	if obj.running {
 		return obj.StopTimer, fmt.Errorf("timer already started")
 	}
-	obj.mutex.Unlock()
+	obj.timer = make(chan struct{})
+	obj.running = true
 	obj.wg.Add(1)
 	go func() {
 		defer obj.wg.Done()
 		for {
 			select {
 			case _, ok := <-obj.timer: // reset signal channel
-				if !ok { // channel is closed
-					return // false to exit
+				if !ok {
+					return
 				}
 				obj.SetConverged(false)
 
@@ -393,8 +460,8 @@ func (obj *cuid) StartTimer() (func() error, error) {
 				obj.SetConverged(true) // converged!
 				select {
 				case _, ok := <-obj.timer: // reset signal channel
-					if !ok { // channel is closed
-						return // false to exit
+					if !ok {
+						return
 					}
 				}
 			}
@@ -403,8 +470,8 @@ func (obj *cuid) StartTimer() (func() error, error) {
 	return obj.StopTimer, nil
 }
 
-// ResetTimer resets the counter to zero if using a StartTimer internally.
-func (obj *cuid) ResetTimer() error {
+// ResetTimer resets the timer to zero.
+func (obj *UID) ResetTimer() error {
 	obj.mutex.Lock()
 	defer obj.mutex.Unlock()
 	if obj.running {
@@ -414,8 +481,8 @@ func (obj *cuid) ResetTimer() error {
 	return fmt.Errorf("timer hasn't been started")
 }
 
-// StopTimer stops the running timer permanently until a StartTimer is run.
-func (obj *cuid) StopTimer() error {
+// StopTimer stops the running timer.
+func (obj *UID) StopTimer() error {
 	obj.mutex.Lock()
 	defer obj.mutex.Unlock()
 	if !obj.running {

converger/converger_test.go

@@ -0,0 +1,43 @@
+// Mgmt
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
+
+//go:build !root
+
+package converger
+
+import (
+	"testing"
+)
+
+func TestBufferedChan1(t *testing.T) {
+	ch := make(chan bool, 1)
+	ch <- true
+	close(ch) // closing a channel that's not empty should not block
+	// must be able to exit without blocking anywhere
+}

debian/copyright

@@ -3,7 +3,7 @@ Upstream-Name: mgmt
 Source: <https://github.com/purpleidea/mgmt>
 
 Files: *
-Copyright: Copyright (C) 2013-2018+ James Shubin and the project contributors
+Copyright: Copyright (C) 2013-2024+ James Shubin and the project contributors
 License: GPL-3.0
 
 License: GPL-3.0
@@ -19,3 +19,15 @@ License: GPL-3.0
 
   You should have received a copy of the GNU General Public License
   along with this program.  If not, see <https://www.gnu.org/licenses/>.
+
+  Additional permission under GNU GPL version 3 section 7
+
+  If you modify this program, or any covered work, by linking or combining it
+  with embedded mcl code and modules (and that the embedded mcl code and
+  modules which link with this program, contain a copy of their source code in
+  the authoritative form) containing parts covered by the terms of any other
+  license, the licensors of this program grant you additional permission to
+  convey the resulting work. Furthermore, the licensors of this program grant
+  the original author, James Shubin, additional permission to update this
+  additional permission if he deems it necessary to achieve the goals of this
+  additional permission.

doc.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,7 +13,19 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 // Package main provides the main entrypoint for using the `mgmt` software.
 package main

docker/Dockerfile

@@ -1,10 +1,10 @@
-FROM golang:1.9
+FROM golang:1.20
 
 MAINTAINER Michał Czeraszkiewicz <contact@czerasz.com>
 
 # Set the reset cache variable
 # Read more here: http://czerasz.com/2014/11/13/docker-tip-and-tricks/#use-refreshedat-variable-for-better-cache-control
-ENV REFRESHED_AT 2017-11-16
+ENV REFRESHED_AT 2020-09-23
 
 # Update the package list to be able to use required packages
 RUN apt-get update

docker/Dockerfile-fedora.build

@@ -0,0 +1,20 @@
+FROM fedora:38
+LABEL org.opencontainers.image.authors="laurent.indermuehle@pm.me"
+
+ENV GOPATH=/root/gopath
+ENV PATH=/root/gopath/bin:/usr/local/sbin:/sbin:/bin:/usr/sbin:/usr/bin:/usr/local/go/bin:/usr/local/bin
+ENV LD_LIBRARY_PATH=/usr/lib64
+ENV PKG_CONFIG_PATH=/usr/lib64/pkgconfig
+
+# This forces make-deps.sh to install Ragel 6.1 instead of 7.0
+ENV DOCKER=true
+
+RUN dnf -y install wget unzip git make which gcc gcc-c++ ruby golang
+
+RUN mkdir -p $GOPATH/src/github.com/purpleidea
+WORKDIR $GOPATH/src/github.com/purpleidea
+RUN git clone --recursive https://github.com/purpleidea/mgmt mgmt
+WORKDIR $GOPATH/src/github.com/purpleidea/mgmt
+RUN make deps
+RUN make build
+CMD ["/bin/bash"]

docker/Dockerfile.build

@@ -6,7 +6,7 @@ ENV PATH=/opt/rh/rh-ruby22/root/usr/bin:/root/gopath/bin:/usr/local/sbin:/sbin:/
 ENV LD_LIBRARY_PATH=/opt/rh/rh-ruby22/root/usr/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
 ENV PKG_CONFIG_PATH=/opt/rh/rh-ruby22/root/usr/lib64/pkgconfig${PKG_CONFIG_PATH:+:${PKG_CONFIG_PATH}}
 
-RUN yum -y install epel-release wget unzip git make which centos-release-scl gcc && sed -i "s/enabled=0/enabled=1/" /etc/yum.repos.d/epel-testing.repo && yum -y install rh-ruby22 && wget -O /opt/go1.9.1.linux-amd64.tar.gz https://storage.googleapis.com/golang/go1.9.1.linux-amd64.tar.gz && tar -C /usr/local -xzf /opt/go1.9.1.linux-amd64.tar.gz
+RUN yum -y install epel-release wget unzip git make which centos-release-scl gcc && sed -i "s/enabled=0/enabled=1/" /etc/yum.repos.d/epel-testing.repo && yum -y install rh-ruby22 && wget -O /opt/go1.20.11.linux-amd64.tar.gz https://storage.googleapis.com/golang/go1.20.11.linux-amd64.tar.gz && tar -C /usr/local -xzf /opt/go1.20.11.linux-amd64.tar.gz
 RUN mkdir -p $GOPATH/src/github.com/purpleidea && cd $GOPATH/src/github.com/purpleidea && git clone --recursive https://github.com/purpleidea/mgmt
 RUN go get -u gopkg.in/alecthomas/gometalinter.v1 && cd $GOPATH/src/github.com/purpleidea/mgmt && make deps && make build
 CMD ["/bin/bash"]

docker/Dockerfile.development

@@ -1,10 +1,10 @@
-FROM golang:1.9
+FROM golang:1.20
 
 MAINTAINER Michał Czeraszkiewicz <contact@czerasz.com>
 
 # Set the reset cache variable
 # Read more here: http://czerasz.com/2014/11/13/docker-tip-and-tricks/#use-refreshedat-variable-for-better-cache-control
-ENV REFRESHED_AT 2017-11-16
+ENV REFRESHED_AT 2019-02-06
 
 RUN apt-get update
 
@@ -27,8 +27,5 @@ WORKDIR /home/$USER_NAME/mgmt
 # Install dependencies
 RUN make deps
 
-# Chown $GOPATH
-RUN chown -R ${USER_ID}:${GROUP_ID} /go
-
 # Change user
 USER ${USER_NAME}

docs/conf.py

@@ -51,7 +51,7 @@ master_doc = 'index'
 
 # General information about the project.
 project = u'mgmt'
-copyright = u'2013-2018+ James Shubin and the project contributors'
+copyright = u'2013-2024+ James Shubin and the project contributors'
 author = u'James Shubin'
 
 # The version info for the project you're documenting, acts as replacement for

docs/development.md

@@ -3,7 +3,119 @@
 This document contains some additional information and help regarding
 developing `mgmt`. Useful tools, conventions, etc.
 
-Be sure to read [quick start guide](docs/quick-start-guide.md) first.
+Be sure to read [quick start guide](quick-start-guide.md) first.
+
+## Vagrant
+
+If you would like to avoid doing the above steps manually, we have prepared a
+[Vagrant](https://www.vagrantup.com/) environment for your convenience. From the
+project directory, run a `vagrant up`, and then a `vagrant status`. From there,
+you can `vagrant ssh` into the `mgmt` machine. The `MOTD` will explain the rest.
+This environment isn't commonly used by the `mgmt` developers, so it might not
+be working properly.
+
+## Using Docker
+
+Alternatively, you can check out the [docker-guide](docker-guide.md) in order to
+develop or deploy using docker. This method is not endorsed or supported, so use
+at your own risk, as it might not be working properly.
+
+## Information about dependencies
+
+Software projects have a few different kinds of dependencies. There are _build_
+dependencies, _runtime_ dependencies, and additionally, a few extra dependencies
+required for running the _test_ suite.
+
+### Build
+
+* `golang` 1.20 or higher (required, available in some distros and distributed
+as a binary officially by [golang.org](https://golang.org/dl/))
+
+### Runtime
+
+A relatively modern GNU/Linux system should be able to run `mgmt` without any
+problems. Since `mgmt` runs as a single statically compiled binary, all of the
+library dependencies are included. It is expected, that certain advanced
+resources require host specific facilities to work. These requirements are
+listed below:
+
+| Resource | Dependency        | Version                     | Check version with                                        |
+|----------|-------------------|-----------------------------|-----------------------------------------------------------|
+| augeas   | augeas-devel      | `augeas 1.6` or greater     | `dnf info augeas-devel` or `apt-cache show libaugeas-dev` |
+| file     | inotify           | `Linux 2.6.27` or greater   | `uname -a`                                                |
+| hostname | systemd-hostnamed | `systemd 25` or greater     | `systemctl --version`                                     |
+| nspawn   | systemd-nspawn    | `systemd ???` or greater    | `systemctl --version`                                     |
+| pkg      | packagekitd       | `packagekit 1.x` or greater | `pkcon --version`                                         |
+| svc      | systemd           | `systemd ???` or greater    | `systemctl --version`                                     |
+| virt     | libvirt-devel     | `libvirt 1.2.0` or greater  | `dnf info libvirt-devel` or `apt-cache show libvirt-dev`  |
+| virt     | libvirtd          | `libvirt 1.2.0` or greater  | `libvirtd --version`                                      |
+
+For building a visual representation of the graph, `graphviz` is required.
+
+To build `mgmt` without augeas support please run:
+`GOTAGS='noaugeas' make build`
+
+To build `mgmt` without libvirt support please run:
+`GOTAGS='novirt' make build`
+
+To build `mgmt` without docker support please run:
+`GOTAGS='nodocker' make build`
+
+To build `mgmt` without augeas, libvirt or docker support please run:
+`GOTAGS='noaugeas novirt nodocker' make build`
+
+## OSX/macOS/Darwin development
+
+Developing and running `mgmt` on macOS is currently not supported (but not
+discouraged either). Meaning it might work but in the case it doesn't you would
+have to provide your own patches to fix problems (the project maintainer and
+community are glad to assist where needed).
+
+There are currently some issues that make `mgmt` less suitable to run for
+provisioning macOS. But as a client to provision remote servers it should run
+fine.
+
+Since the primary supported systems are Linux and these are the environments
+tested, it is wise to run these suites during macOS development as well. To ease
+this, Docker can be leveraged ([Docker for Mac](https://docs.docker.com/docker-for-mac/)).
+
+Before running any of the commands below create the development Docker image:
+
+```
+docker/scripts/build-development
+```
+
+This image requires updating every time dependencies (`make-deps.sh`) changes.
+
+Then to run the test suite:
+
+```
+docker run --rm -ti \
+	-v $PWD:/go/src/github.com/purpleidea/mgmt/ \
+	-w /go/src/github.com/purpleidea/mgmt/ \
+	purpleidea/mgmt:development \
+	make test
+```
+
+For convenience this command is wrapped in `docker/scripts/exec-development`.
+
+Basically any command can be executed this way. Because the repository source is
+mounted into the Docker container invocation will be quick and allow rapid
+testing, for example:
+
+```
+docker/scripts/exec-development test/test-shell.sh load0.sh
+```
+
+Other examples:
+
+```
+docker/scripts/exec-development make build
+docker/scripts/exec-development ./mgmt run --tmp-prefix lang examples/lang/load0.mcl
+```
+
+Be advised that this method is not supported and it might not be working
+properly.
 
 ## Testing
 
@@ -45,5 +157,6 @@ individual tests to run.
 
 ### IDE/Editor support
 
-- Emacs: see `misc/emacs/`
-- [Textmate](https://github.com/aequitas/mgmt.tmbundle)
+* Emacs: see `misc/emacs/`
+* [Textmate](https://github.com/aequitas/mgmt.tmbundle)
+* [VSCode](https://github.com/aequitas/mgmt.vscode)

docs/documentation.md

@@ -122,38 +122,15 @@ entire set of running mgmt agents will need to all simultaneously converge for
 the group to exit. This is particularly useful for bootstrapping new clusters
 which need to exchange information that is only available at run time.
 
+This existed in earlier versions of mgmt as a `--remote` option, but it has been
+removed and is being ported to a more powerful variant where you can remote
+execute via a `remote` resource.
+
 #### Blog post
 
 You can read the introductory blog post about this topic here:
 [https://purpleidea.com/blog/2016/10/07/remote-execution-in-mgmt/](https://purpleidea.com/blog/2016/10/07/remote-execution-in-mgmt/)
 
-### Puppet support
-
-You can supply a Puppet manifest instead of creating the (YAML) graph manually.
-Puppet must be installed and in `mgmt`'s search path. You also need the
-[ffrank-mgmtgraph Puppet module](https://forge.puppet.com/ffrank/mgmtgraph).
-
-Invoke `mgmt` with the `--puppet` switch, which supports 3 variants:
-
-1. Request the configuration from the Puppet Master (like `puppet agent` does)
-
-	`mgmt run --puppet agent`
-
-2. Compile a local manifest file (like `puppet apply`)
-
-	`mgmt run --puppet /path/to/my/manifest.pp`
-
-3. Compile an ad hoc manifest from the commandline (like `puppet apply -e`)
-
-	`mgmt run --puppet 'file { "/etc/ntp.conf": ensure => file }'`
-
-For more details and caveats see [Puppet.md](Puppet.md).
-
-#### Blog post
-
-An introductory post on the Puppet support is on
-[Felix's blog](http://ffrank.github.io/features/2016/06/19/puppet-powered-mgmt/).
-
 ## Reference
 
 Please note that there are a number of undocumented options. For more
@@ -164,6 +141,7 @@ If you feel that a well used option needs documenting here, please patch it!
 ### Overview of reference
 
 * [Meta parameters](#meta-parameters): List of available resource meta parameters.
+* [Lang metadata file](#lang-metadata-file): Lang metadata file format.
 * [Graph definition file](#graph-definition-file): Main graph definition file.
 * [Command line](#command-line): Command line parameters.
 * [Compilation options](#compilation-options): Compilation options.
@@ -249,21 +227,91 @@ integer, then that value is the max size for that semaphore. Valid semaphore
 id's include: `some_id`, `hello:42`, `not:smart:4` and `:13`. It is expected
 that the last bare example be only used by the engine to add a global semaphore.
 
+#### Rewatch
+
+Boolean. Rewatch specifies whether we re-run the Watch worker during a graph
+swap if it has errored. When doing a graph compare to swap the graphs, if this
+is true, and this particular worker has errored, then we'll remove it and add it
+back as a new vertex, thus causing it to run again. This is different from the
+`Retry` metaparam which applies during the normal execution. It is only when
+this is exhausted that we're in permanent worker failure, and only then can we
+rely on this metaparam.
+
+#### Realize
+
+Boolean. Realize ensures that the resource is guaranteed to converge at least
+once before a potential graph swap removes or changes it. This guarantee is
+useful for fast changing graphs, to ensure that the brief creation of a resource
+is seen. This guarantee does not prevent against the engine quitting normally,
+and it can't guarantee it if the resource is blocked because of a failed
+pre-requisite resource.
+*XXX: This is currently not implemented!*
+
+#### Reverse
+
+Boolean. Reverse is a property that some resources can implement that specifies
+that some "reverse" operation should happen when that resource "disappears". A
+disappearance happens when a resource is defined in one instance of the graph,
+and is gone in the subsequent one. This disappearance can happen if it was
+previously in an if statement that then becomes false.
+
+This is helpful for building robust programs with the engine. The engine adds a
+"reversed" resource to that subsequent graph to accomplish the desired "reverse"
+mechanics. The specifics of what this entails is a property of the particular
+resource that is being "reversed".
+
+It might be wise to combine the use of this meta parameter with the use of the
+`realize` meta parameter to ensure that your reversed resource actually runs at
+least once, if there's a chance that it might be gone for a while.
+
+### Lang metadata file
+
+Any module *must* have a metadata file in its root. It must be named
+`metadata.yaml`, even if it's empty. You can specify zero or more values in yaml
+format which can change how your module behaves, and where the `mcl` language
+looks for code and other files. The most important top level keys are: `main`,
+`path`, `files`, and `license`.
+
+#### Main
+
+The `main` key points to the default entry point of your code. It must be a
+relative path if specified. If it's empty it defaults to `main.mcl`. It should
+generally not be changed. It is sometimes set to `main/main.mcl` if you'd like
+your modules code out of the root and into a child directory for cases where you
+don't plan on having a lot deeper imports relative to `main.mcl` and all those
+files would clutter things up.
+
+#### Path
+
+The `path` key specifies the modules import search directory to use for this
+module. You can specify this if you'd like to vendor something for your module.
+In general, if you use it, please use the convention: `path/`. If it's not
+specified, you will default to the parent modules directory.
+
+#### Files
+
+The `files` key specifies some additional files that will get included in your
+deploy. It defaults to `files/`.
+
+#### License
+
+The `license` key allows you to specify a license for the module. Please specify
+one so that everyone can enjoy your code! Use a "short license identifier", like
+`LGPLv3+`, or `MIT`. The former is a safe choice if you're not sure what to use.
+
 ### Graph definition file
 
 graph.yaml is the compiled graph definition file. The format is currently
-undocumented, but by looking through the [examples/](https://github.com/purpleidea/mgmt/tree/master/examples)
-you can probably figure out most of it, as it's fairly intuitive.
+undocumented, but by looking through the [examples/](https://github.com/purpleidea/mgmt/tree/master/examples/yaml/)
+you can probably figure out most of it, as it's fairly intuitive. It's not
+recommended that you use this, since it's preferable to write code in the
+[mcl language](language-guide.md) front-end.
 
 ### Command line
 
 The main interface to the `mgmt` tool is the command line. For the most recent
 documentation, please run `mgmt --help`.
 
-#### `--yaml <graph.yaml>`
-
-Point to a graph file to run.
-
 #### `--converged-timeout <seconds>`
 
 Exit if the machine has converged for approximately this many seconds.
@@ -289,33 +337,11 @@ collision with this globally defined semaphore. The size value must be greater
 than zero at this time. The traditional non-parallel execution found in config
 management tools such as `Puppet` can be obtained with `--sema 1`.
 
-#### `--remote <graph.yaml>`
-
-Point to a graph file to run on the remote host specified within. This parameter
-can be used multiple times if you'd like to remotely run on multiple hosts in
-parallel.
-
-#### `--allow-interactive`
-
-Allow interactive prompting for SSH passwords if there is no authentication
-method that works.
-
 #### `--ssh-priv-id-rsa`
 
 Specify the path for finding SSH keys. This defaults to `~/.ssh/id_rsa`. To
 never use this method of authentication, set this to the empty string.
 
-#### `--cconns`
-
-The maximum number of concurrent remote ssh connections to run. This defaults
-to `0`, which means unlimited.
-
-#### `--no-caching`
-
-Don't allow remote caching of the remote execution binary. This will require
-the binary to be copied over for every remote execution, but it limits the
-likelihood that there is leftover information from the configuration process.
-
 #### `--prefix <path>`
 
 Specify a path to a custom working directory prefix. This directory will get
@@ -333,8 +359,8 @@ default prefix. This can't be combined with the `--prefix` option.
 If this option is specified, we will attempt to fall back to a temporary prefix
 if the primary prefix couldn't be created. This is useful for avoiding failures
 in environments where the primary prefix may or may not be available, but you'd
-like to try. The canonical example is when running `mgmt` with `--remote` there
-might be a cached copy of the binary in the primary prefix, but in case there's
+like to try. The canonical example is when running `mgmt` with remote execution
+there might be a cached copy of the binary in the primary prefix, but if there's
 no binary available continue working in a temporary directory to avoid failure.
 
 ### Compilation options
@@ -417,7 +443,7 @@ To report any bugs, please file a ticket at: [https://github.com/purpleidea/mgmt
 
 ## Authors
 
-Copyright (C) 2013-2018+ James Shubin and the project contributors
+Copyright (C) 2013-2024+ James Shubin and the project contributors
 
 Please see the
 [AUTHORS](https://github.com/purpleidea/mgmt/tree/master/AUTHORS) file

docs/faq.md

@@ -9,6 +9,18 @@ I wanted a next generation config management solution that didn't have all of
 the design flaws or limitations that the current generation of tools do, and no
 tool existed!
 
+### Why did you choose `golang` for the project?
+
+When I started working on the project, I needed to choose a language that
+already had an implementation of a distributed consensus algorithm available.
+That meant [Paxos](https://en.wikipedia.org/wiki/Paxos_(computer_science)) or
+[Raft](https://en.wikipedia.org/wiki/Raft_(computer_science)). Golang was one
+language that actually had two different Raft implementations, `etcd`, and
+`consul`. Other design requirements included something that was reasonably fast,
+typed and memory-safe, and suited for systems engineering. After a reasonably
+extensive search, I chose `golang`. I think it was the right decision. There are
+a number of other features of the language which helped influence the decision.
+
 ### How do I contribute to the project if I don't know `golang`?
 
 There are many different ways you can contribute to the project. They can be
@@ -41,10 +53,11 @@ find a number of tutorials online.
 3. Spend between four to six hours with the [golang tour](https://tour.golang.org/).
 Skip over the longer problems, but try and get a solid overview of everything.
 If you forget something, you can always go back and repeat those parts.
-4. Connect to our [#mgmtconfig](https://webchat.freenode.net/?channels=#mgmtconfig)
-IRC channel on the [Freenode](https://freenode.net/) network. You can use any
-IRC client that you'd like, but the [hosted web portal](https://webchat.freenode.net/?channels=#mgmtconfig)
-will suffice if you don't know what else to use.
+4. Connect to our [#mgmtconfig](https://web.libera.chat/?channels=#mgmtconfig)
+IRC channel on the [Libera.Chat](https://libera.chat/) network. You can use any
+IRC client that you'd like, but the [hosted web portal](https://web.libera.chat/?channels=#mgmtconfig)
+will suffice if you don't know what else to use. [Here are a few suggestions for
+alternative clients.](https://libera.chat/guides/clients)
 5. Now it's time to try and starting writing a patch! We have tagged a bunch of
 [open issues as #mgmtlove](https://github.com/purpleidea/mgmt/issues?q=is%3Aissue+is%3Aopen+label%3Amgmtlove)
 for new users to have somewhere to get involved. Look through them to see if
@@ -57,6 +70,8 @@ hacking!
 
 ### Is this project ready for production?
 
+It's getting pretty close. I'm able to write modules for it now!
+
 Compared to some existing automation tools out there, mgmt is a relatively new
 project. It is probably not as feature complete as some other software, but it
 also offers a number of features which are not currently available elsewhere.
@@ -123,6 +138,58 @@ The downside to this approach is that you won't benefit from the automatic
 elastic nature of the embedded etcd servers, and that you're responsible if you
 accidentally break your etcd cluster, or if you use an unsupported version.
 
+### In `mgmt` you talk about events. What is this referring to?
+
+Mgmt has two main concepts that involve "events":
+1. Events in the [resource primitive](resource-guide.md).
+2. Events in the [reactive language](language-guide.md).
+
+Each resource primitive in mgmt can test (check) and set (apply) the desired
+state that was requested of it. This is familiar to what is common with existing
+tools such as `Puppet`, `Ansible`, `Chef`, `Terraform`, etc... In addition,
+`mgmt` can also **watch** the state and detect changes. As a result, it never
+has to waste time and cpu resources by polling to test and set state, leading to
+a design which is algorithmically much faster than the existing generation of
+tools.
+
+To describe the set of resources to apply, mgmt describes this collection with a
+language. In order to model the time component of infrastructure, we use a
+special kind of language called an [FRP](https://en.wikipedia.org/wiki/Functional_reactive_programming).
+This language has a built-in concept that we call "events", and which means that
+we re-evaluate the relevant portions of the code whenever a value or function
+has an event that tells us that it changed. The `R` in `FRP` stands for
+reactive. This is similar to how a spreadsheet updates dependent cells when a
+pre-requisite value is modified. [This article](https://en.wikipedia.org/wiki/Reactive_programming)
+provides a bit more background.
+
+Whenever any of the streams of values in the language change, the program is
+partially re-evaluated. The output of any mgmt program is a [DAG](https://en.wikipedia.org/wiki/Directed_acyclic_graph)
+of resources, or more precisely, a stream of resource graphs. Since we have
+events per-resource, we can efficiently switch from one desired-state resource
+graph to the next without re-checking their individual states, since we've been
+monitoring them all along.
+
+One side-effect of all this, is that if a rogue systems administrator manually
+changes the state of any managed resource, mgmt will detect this and attempt to
+revert the change. This makes for excellent live demos, but is not the primary
+design goal. It is a consequence of tracking state so that graph changes are
+efficient. We implement the event detection via an intentional per-resource
+[main loop](https://en.wikipedia.org/wiki/Event_loop) which can enable other
+interesting functionality too!
+
+Make sure to get rid of your rogue sysadmin! ;)
+
+### Do I need to run `mgmt` as `root`?
+
+No and yes. It depends. Nothing in mgmt explicitly requires root in the design,
+however mgmt will require root only if the changes to your system that you want
+it to make require root.
+
+For example, if you use it to manage files that require root access to modify,
+then you'll need root. If you only use it to manage files and resources
+elsewhere, then it shouldn't need root. Many resources are perfectly usable
+without root, and virtually all of my live demos are done without root.
+
 ### How can I run `mgmt` on-demand, or in `cron`, instead of continuously?
 
 By default, `mgmt` will run continuously in an attempt to keep your machine in a
@@ -146,42 +213,171 @@ requires a number of seconds as an argument.
 #### Example:
 
 ```
-./mgmt run --lang examples/lang/hello0.mcl --converged-timeout=5
+./mgmt run lang examples/lang/hello0.mcl --converged-timeout=5
 ```
 
-### What does the error message about an inconsistent dataDir mean?
+### Can I run `mgmt` for type-checking only?
+
+Yes, you can, add the `--only-unify` option to the lang frontend while using the
+`run` command, and it will exit after type unification.
+
+#### Example:
+
+```
+./mgmt run --tmp-prefix lang --only-unify examples/lang/hello0.mcl
+```
+
+It will also print how long it took on either success or failure. Keep in mind
+that even if you pass type unification, an `mgmt` run can still fail later on
+for other reasons, although these are mostly runtime considerations.
+
+### Why is type unification happening twice with `mgmt run`?
+
+When you use the `run` action, it runs all of the compile time checks (including
+type unification) that are possible, and then packages everything up into a
+deployable object and runs the same mechanism that `mgmt deploy` uses, sending
+the deploy to itself. At this point, mgmt starts up as a server, and receives
+the deploy. It will then need to run type unification again before running the
+code.
+
+You can skip the first type unification check by adding the `--skip-unify`
+option to the lang frontend when using the `run` command.
+
+You can also skip this check when running the `deploy` action, but if your code
+doesn't pass, you might be deploying broken code. This is not recommended.
+
+#### Example:
+
+```
+./mgmt run --tmp-prefix lang --skip-unify examples/lang/hello0.mcl
+```
+
+### Why does my file resource error with `no such file or directory`?
+
+If you create a file resource and only specify the content like this:
+
+```
+file "/tmp/foo" {
+	content => "hello world\n",
+}
+```
+
+Then this will attempt to set the contents of that file to the desired string,
+but *only* if that file already exists. If you'd like to ensure that it also
+gets created in case it is not present, then you must also specify the state:
+
+```
+file "/tmp/foo" {
+	state => $const.res.file.state.exists,
+	content => "hello world\n",
+}
+```
+
+Similar logic applies for situations when you only specify the `mode` parameter.
+
+This all turns out to be more safe and "correct", in that it would error and
+prevent masking an error for a situation when you expected a file to already be
+at that location. It also turns out to simplify the internals significantly, and
+remove an ambiguous scenario with the reversible file resource.
+
+### Why do function names inside of templates include underscores?
+
+The golang template library which we use to implement the template() function
+doesn't support the dot notation, so we import all our normal functions, and
+just replace dots with underscores. As an example, the standard `datetime.print`
+function is shown within mcl scripts as datetime_print after being imported.
+
+### On startup `mgmt` hangs after: `etcd: server: starting...`.
 
 If you get an error message similar to:
 
 ```
-Etcd: Connect: CtxError...
-Etcd: CtxError: Reason: CtxDelayErr(5s): No endpoints available yet!
-Etcd: Connect: Endpoints: []
-Etcd: The dataDir (/var/lib/mgmt/etcd) might be inconsistent or corrupt.
+etcd: server: starting...
+etcd: server: start timeout of 1m0s reached
+etcd: server: close timeout of 15s reached
 ```
 
-This happens when there are a series of fatal connect errors in a row. This can
-happen when you start `mgmt` using a dataDir that doesn't correspond to the
-current cluster view. As a result, the embedded etcd server never finishes
-starting up, and as a result, a default endpoint never gets added. The solution
-is to either reconcile the mistake, and if there is no important data saved, you
-can remove the etcd dataDir. This is typically `/var/lib/mgmt/etcd/member/`.
+But nothing happens afterwards, this can be due to a corrupt etcd storage
+directory. Each etcd server embedded in mgmt must have a special directory where
+it stores local state. It must not be shared by more than one individual member.
+This dir is typically `/var/lib/mgmt/etcd/member/`. If you accidentally use it
+(for example during testing) with a different cluster view, then you can corrupt
+it. This can happen if you use it with more than one different hostname.
 
-### Why do resources have both a `Compare` method and an `IFF` (on the UID) method?
+The solution is to avoid making this mistake, and if there is no important data
+saved, you can remove the etcd member dir and start over.
 
-The `Compare()` methods are for determining if two resources are effectively the
-same, which is used to make graph change delta's efficient. This is when we want
-to change from the current running graph to a new graph, but preserve the common
-vertices. Since we want to make this process efficient, we only update the parts
-that are different, and leave everything else alone. This `Compare()` method can
-tell us if two resources are the same.
+### On running `make` to build a new version, it errors with: `Text file busy`.
 
-The `IFF()` method is part of the whole UID system, which is for discerning if a
-resource meets the requirements another expects for an automatic edge. This is
-because the automatic edge system assumes a unified UID pattern to test for
-equality. In the future it might be helpful or sane to merge the two similar
-comparison functions although for now they are separate because they are
-actually answer different questions.
+If you get an error like:
+
+```
+cp: cannot create regular file 'mgmt': Text file busy
+```
+
+This can happen if you ran `make build` (or just `make`) when there was already
+an instance of mgmt running, or if a related file locking issue occurred. To
+solve this, shutdown and running mgmt process, run `rm mgmt` to remove the file,
+and then get a new one by running `make` again.
+
+### Type unification error: "could not unify types: 2 unconsumed generators".
+
+Look carefully at the following code:
+
+```
+$num = 42
+print "hello" {
+	msg => "My favourite number is ${num}",
+}
+```
+
+What's actually happening is that we can't unify `str + int`, because our
+addition operator only supports `str + str`, `int + int`, and `float + float`.
+The string interpolation feature can only combine strings. One solution is to
+instead write:
+
+```
+$num = "42" # now this is a string
+print "hello" {
+	msg => "My favourite number is ${num}",
+}
+```
+
+Yes we know the compiler gives horrible error messages, and yes we would
+absolutely love your help improving this.
+
+### The run and deploy commands don't parse correctly when used with `--seeds`.
+
+If you're running a command with `--seeds`, `--server-urls`, or `--client-urls`,
+then make sure you are using an equals sign between the flag name and the value.
+For example, if you were to run:
+
+```
+# wrong invocation!
+mgmt deploy --no-git --seeds http://127.0.0.1:2379 lang code/test.mcl
+```
+
+Then the `--seeds` flag would interpret `lang` and `code/test.mcl` as additional
+seeds. This flag as well as the other aforementioned ones all accept multiple
+values. Use an equals sign to guarantee you enter the correct data, eg:
+
+```
+# better invocation! (note the equals sign)
+mgmt deploy --no-git --seeds=http://127.0.0.1:2379 lang code/test.mcl
+```
+
+This is caused by a parsing peculiarity of the CLI library that we are using.
+This is tracked upstream at: [https://github.com/alexflint/go-arg/issues/239](https://github.com/alexflint/go-arg/issues/239).
+We have a workaround in place to mitigate it and attempt to show you a helpful
+error message, but it's also documented here in the meantime. The error you will
+see is: `cli parse error: missing equals sign for list element`.
+
+### The docs speaks of `--remote` but the CLI errors out?
+
+The `--remote` flag existed in an earlier version of mgmt. It was removed and
+will be replaced with a more powerful version, which is a "remote" resource. The
+code is mostly ready but it's not finished. If you'd like to help finish it or
+sponsor the work, please let me know.
 
 ### Does this support Windows? OSX? GNU Hurd?
 
@@ -190,7 +386,7 @@ serious automation workloads. Support for non-Linux operating systems isn't a
 high priority of mine, but we're happy to accept patches for missing features
 or resources that you think would make sense on your favourite platform.
 
-### Why aren't you using `glide` or `godep` for dependency management?
+### Why aren't you using `glide`, `godep` or `go mod` for dependency management?
 
 Vendoring dependencies means that as the git master branch of each dependency
 marches on, you're left behind using an old version. As a result, bug fixes and
@@ -245,6 +441,14 @@ Don't blindly use the tools that others tell you to. Learn what they do, think
 for yourself, and become a power user today! That process led us to using
 `git submodules`. Hopefully you'll come to the same conclusions that we did.
 
+**UPDATE:**
+
+After golang made it virtually impossible to build without `go.mod` stuff, we've
+switched to it since golang 1.16. I still think the above approach was better,
+and that the `go mod` tooling should have been a layer on top of git submodules
+so that we don't grow yet another lock file format, and existing folks who are
+comfortable with `git` can use those tools directly.
+
 ### Did you know that there is a band named `MGMT`?
 
 I didn't realize this when naming the project, and it is accidental. After much
@@ -258,10 +462,9 @@ which definitely existed before the band did.
 
 ### You didn't answer my question, or I have a question!
 
-It's best to ask on [IRC](https://webchat.freenode.net/?channels=#mgmtconfig)
-to see if someone can help you. Once we get a big enough community going, we'll
-add a mailing list. If you don't get any response from the above, you can
-contact me through my [technical blog](https://purpleidea.com/contact/)
-and I'll do my best to help. If you have a good question, please add it as a
-patch to this documentation. I'll merge your question, and add a patch with the
-answer!
+It's best to ask on [IRC](https://web.libera.chat/?channels=#mgmtconfig)
+to see if someone can help you. If you don't get a response from IRC, you can
+contact me through my [technical blog](https://purpleidea.com/contact/) and I'll
+do my best to help. If you have a good question, please add it as a patch to
+this documentation. I'll merge your question, and add a patch with the answer!
+For news and updates, subscribe to the [mailing list](https://www.redhat.com/mailman/listinfo/mgmtconfig-list).

docs/function-guide.md

@@ -37,10 +37,12 @@ available types and values in the mgmt language. It is very easy to use, and
 should be fairly intuitive. Most of what you'll need to know can be inferred
 from looking at example code.
 
-To implement a function, you'll need to create a file in
-[`lang/funcs/simple/`](https://github.com/purpleidea/mgmt/tree/master/lang/funcs/simple/).
-The function should be implemented as a `FuncValue` in our type system. It is
-then registered with the engine during `init()`. An example explains it best:
+To implement a function, you'll need to create a file that imports the
+[`lang/funcs/simple/`](https://github.com/purpleidea/mgmt/tree/master/lang/funcs/simple/)
+module. It should probably get created in the correct directory inside of:
+[`lang/core/`](https://github.com/purpleidea/mgmt/tree/master/lang/core/). The
+function should be implemented as a `FuncValue` in our type system. It is then
+registered with the engine during `init()`. An example explains it best:
 
 ### Example
 
@@ -50,14 +52,15 @@ package simple
 import (
 	"fmt"
 
+	"github.com/purpleidea/mgmt/lang/funcs/simple"
 	"github.com/purpleidea/mgmt/lang/types"
 )
 
 // you must register your functions in init when the program starts up
 func init() {
 	// Example function that squares an int and prints out answer as an str.
-	Register("talkingsquare", &types.FuncValue{
-		T: types.NewType("func(a int) str"), // declare the signature
+	simple.ModuleRegister(ModuleName, "talkingsquare", &types.FuncValue{
+		T: types.NewType("func(int) str"), // declare the signature
 		V: func(input []types.Value) (types.Value, error) {
 			i := input[0].Int() // get first arg as an int64
 			// must return the above specified value
@@ -109,31 +112,37 @@ As with the simple, non-polymorphic API, you can only implement [pure](https://e
 functions, without writing too much boilerplate code. They will be automatically
 re-evaluated as needed when their input values change.
 
-To implement a function, you'll need to create a file in
-[`lang/funcs/simplepoly/`](https://github.com/purpleidea/mgmt/tree/master/lang/funcs/simplepoly/).
-The function should be implemented as a list of `FuncValue`'s in our type
-system. It is then registered with the engine during `init()`. You may also use
-the `variant` type in your type definitions. This special type will never be
-seen inside a running program, and will get converted to a concrete type if a
+To implement a function, you'll need to create a file that imports the
+[`lang/funcs/simplepoly/`](https://github.com/purpleidea/mgmt/tree/master/lang/funcs/simplepoly/)
+module. It should probably get created in the correct directory inside of:
+[`lang/core/`](https://github.com/purpleidea/mgmt/tree/master/lang/core/). The
+function should be implemented as a list of `FuncValue`'s in our type system. It
+is then registered with the engine during `init()`. You may also use the
+`variant` type in your type definitions. This special type will never be seen
+inside a running program, and will get converted to a concrete type if a
 suitable match to this signature can be found. Be warned that signatures which
 contain too many variants, or which are very general, might be hard for the
-compiler to match, and ambiguous type graphs make for user compiler errors.
+compiler to match, and ambiguous type graphs make for user compiler errors. The
+top-level type must still be a function type, it may only contain variants as
+part of its signature. It is probably more difficult to unify a function if its
+return type is a variant, as opposed to if one of its args was.
 
 An example explains it best:
 
 ### Example
 
 ```golang
-package simplepoly
-
 import (
 	"fmt"
 
+	"github.com/purpleidea/mgmt/lang/funcs/simplepoly"
 	"github.com/purpleidea/mgmt/lang/types"
 )
 
 func init() {
-	Register("len", []*types.FuncValue{
+	// You may use the simplepoly.ModuleRegister method to register your
+	// function if it's in a module, as seen in the simple function example.
+	simplepoly.Register("len", []*types.FuncValue{
 		{
 			T: types.NewType("func([]variant) int"),
 			V: Len,
@@ -191,7 +200,7 @@ if it meets your needs. Most functions will be able to use that API. If you
 really need something more powerful, then you can use the regular function API.
 What follows are each of the method signatures and a description of each.
 
-### Default
+### Info
 
 ```golang
 Info() *interfaces.Info
@@ -220,7 +229,7 @@ Init(init *interfaces.Init) error
 
 This is called to initialize the function. If something goes wrong, it should
 return an error. It is passed a struct that contains all the important
-information and poiinters that it might need to work with throughout its
+information and pointers that it might need to work with throughout its
 lifetime. As a result, it will need to save a copy to that pointer for future
 use in the other methods.
 
@@ -230,27 +239,6 @@ use in the other methods.
 // Init runs some startup code for this function.
 func (obj *FooFunc) Init(init *interfaces.Init) error {
 	obj.init = init
-	obj.closeChan = make(chan struct{}) // shutdown signal
-	return nil
-}
-```
-
-### Close
-
-```golang
-Close() error
-```
-
-This is called to cleanup the function. It usually causes the stream to
-shutdown. Even if `Stream()` decided to shutdown early, it might still get
-called. It is usually called by the engine to tell the function to shutdown.
-
-#### Example
-
-```golang
-// Close runs some shutdown code for this function and turns off the stream.
-func (obj *FooFunc) Close() error {
-	close(obj.closeChan) // send a signal to tell the stream to close
 	return nil
 }
 ```
@@ -258,23 +246,24 @@ func (obj *FooFunc) Close() error {
 ### Stream
 
 ```golang
-Stream() error
+Stream(context.Context) error
 ```
 
 `Stream` is where the real _work_ is done. This method is started by the
 language function engine. It will run this function while simultaneously sending
-it values on the `input` channel. It will only send a complete set of input
+it values on the `Input` channel. It will only send a complete set of input
 values. You should send a value to the output channel when you have decided that
 one should be produced. Make sure to only use input values of the expected type
 as declared in the `Info` struct, and send values of the similarly declared
 appropriate return type. Failure to do so will may result in a panic and
-sadness.
+sadness. You must shutdown if the input context cancels. You must close the
+`Output` channel if you are done generating new values and/or when you shutdown.
 
 #### Example
 
 ```golang
 // Stream returns the single value that was generated and then closes.
-func (obj *FooFunc) Stream() error {
+func (obj *FooFunc) Stream(ctx context.Context) error {
 	defer close(obj.init.Output) // the sender closes
 	var result string
 	for {
@@ -291,7 +280,7 @@ func (obj *FooFunc) Stream() error {
 
 			result = fmt.Sprintf("the input is: %d", ix)
 
-		case <-obj.closeChan:
+		case <-ctx.Done():
 			return nil
 		}
 
@@ -300,7 +289,7 @@ func (obj *FooFunc) Stream() error {
 			V: result,
 		}:
 
-		case <-obj.closeChan:
+		case <-ctx.Done():
 			return nil
 		}
 	}
@@ -331,8 +320,6 @@ type FooFunc struct {
 	init *interfaces.Init
 
 	// this space can be used if needed
-
-	closeChan chan struct{} // shutdown signal
 }
 ```
 
@@ -343,11 +330,23 @@ also ensures they can be encoded and decoded. Make sure to include the following
 code snippet for this to work.
 
 ```golang
+import "github.com/purpleidea/mgmt/lang/funcs"
+
 func init() { // special golang method that runs once
 	funcs.Register("foo", func() interfaces.Func { return &FooFunc{} })
 }
 ```
 
+Functions inside of built-in modules will need to use the `ModuleRegister`
+method instead.
+
+```golang
+// moduleName is already set to "math" by the math package. Do this in `init`.
+funcs.ModuleRegister(moduleName, "cos", func() interfaces.Func {
+	return &CosFunc{}
+})
+```
+
 ### Composite functions
 
 Composite functions are functions which import one or more existing functions.
@@ -368,9 +367,9 @@ might be different ways you would want to call `printf`, such as:
 `printf("the %s is %d", "answer", 42)` or `printf("3 * 2 = %d", 3 * 2)`. Since
 you couldn't implement the infinite number of possible signatures, this API lets
 you write code which can be coerced into different forms. This makes
-implementing what would appear to be generic or polymorphic, instead something
-that is actually static and that still has the static type safety properties
-that were guaranteed by the mgmt language.
+implementing what would appear to be generic or polymorphic, instead of
+something that is actually static and that still has the static type safety
+properties that were guaranteed by the mgmt language.
 
 Since this is an advanced topic, it is not described in full at this time. For
 more information please have a look at the source code comments, some of the
@@ -426,6 +425,11 @@ generator to build your `FuncValue` implementations, and pass in the unique
 signature to each one as you are building them. Using a generator is a common
 technique which was mentioned previously.
 
+One obvious situation where this might occur is if your function doesn't take
+any inputs! An example `math.fortytwo()` function was implemented that
+demonstrates the use of function generators to pass the type signatures into the
+implementations.
+
 ### Where can I find more information about mgmt?
 
 Additional blog posts, videos and other material [is available!](https://github.com/purpleidea/mgmt/blob/master/docs/on-the-web.md).

docs/index.rst

@@ -14,4 +14,3 @@ Welcome to mgmt's documentation!
    quick-start-guide
    resource-guide
    prometheus
-   puppet-guide

docs/language-guide.md

@@ -54,7 +54,7 @@ can be impossible to infer the item's type.
 
 An unordered set of unique keys of the same type and corresponding value pairs
 of another type, eg:
-`{"boiling" => 100, "freezing" => 0, "room" => "25", "house" => 22, "canada" => -30,}`.
+`{"boiling" => 100, "freezing" => 0, "room" => 25, "house" => 22, "canada" => -30,}`.
 That is to say, all of the keys must have the same type, and all of the values
 must have the same type. You can use any type for either, although it is
 probably advisable to avoid using very complex types as map keys.
@@ -140,6 +140,31 @@ expression
 	include bar("world", 13) # an include can be called multiple times
 	```
 
+- **import**: import a particular scope from this location at a given namespace
+
+	```mcl
+	# a system module import
+	import "fmt"
+
+	# a local, single file import (relative path, not a module)
+	import "dir1/file.mcl"
+
+	# a local, module import (relative path, contents are a module)
+	import "dir2/"
+
+	# a remote module import (absolute remote path, contents are a module)
+	import "git://github.com/purpleidea/mgmt-example1/"
+	```
+
+	or
+
+	```mcl
+	import "fmt" as *	# contents namespaced into top-level names
+	import "foo.mcl"	# namespaced as foo
+	import "dir1/" as bar	# namespaced as bar
+	import "git://github.com/purpleidea/mgmt-example1/"	# namespaced as example1
+	```
+
 All statements produce _output_. Output consists of between zero and more
 `edges` and `resources`. A resource statement can produce a resource, whereas an
 `if` statement produces whatever the chosen branch produces. Ultimately the goal
@@ -165,6 +190,8 @@ resource to control how it behaves. For example, setting the `content` parameter
 of a `file` resource to the string `hello`, will cause the contents of that file
 to contain the string `hello` after it has run.
 
+##### Undefined parameters
+
 For some parameters, there is a distinction between an unspecified parameter,
 and a parameter with a `zero` value. For example, for the file resource, you
 might choose to set the `content` parameter to be the empty string, which would
@@ -179,7 +206,7 @@ value to use if that boolean is true. You can do this with the resource-specific
 $b = true # change me to false and then try editing the file manually
 file "/tmp/mgmt-elvis" {
 	content => $b ?: "hello world\n",
-	state => "exists",
+	state => $const.res.file.state.exists,
 }
 ```
 
@@ -189,6 +216,75 @@ it evaluates to `true`, then the parameter will be used. If no `elvis` operator
 is specified, then the parameter value will also be used. If the parameter is
 not specified, then it will obviously not be used.
 
+##### Meta parameters
+
+Resources may specify meta parameters. To do so, you must add them as you would
+a regular parameter, except that they start with `Meta` and are capitalized. Eg:
+
+```mcl
+file "/tmp/f1" {
+	content => "hello!\n",
+
+	Meta:noop => true,
+	Meta:delay => $b ?: 42,
+	Meta:autoedge => false,
+}
+```
+
+As you can see, they also support the elvis operator, and you can add as many as
+you like. While it is not recommended to add the same meta parameter more than
+once, it does not currently cause an error, and even though the result of doing
+so is officially undefined, it will currently take the last specified value.
+
+You may also specify a single meta parameter struct. This is useful if you'd
+like to reuse a value, or build a combined value programmatically. For example:
+
+```mcl
+file "/tmp/f1" {
+	content => "hello!\n",
+
+	Meta => $b ?: struct{
+		noop => false,
+		retry => -1,
+		delay => 0,
+		poll => 5,
+		limit => 4.2,
+		burst => 3,
+		sema => ["foo:1", "bar:3",],
+		autoedge => true,
+		autogroup => false,
+	},
+}
+```
+
+Remember that the top-level `Meta` field supports the elvis operator, while the
+individual struct fields in the struct type do not. This is to be expected, but
+since they are syntactically similar, it is worth mentioning to avoid confusion.
+
+Please note that at the moment, you must specify a full metaparams struct, since
+partial struct types are currently not supported in the language. Patches are
+welcome if you'd like to add this tricky feature!
+
+##### Resource naming
+
+Each resource must have a unique name of type `str` that is used to uniquely
+identify that resource, and can be used in the functioning of the resource at
+that resources discretion. For example, the `file` resource uses the unique name
+value to specify the path.
+
+Alternatively, the name value may be a list of strings `[]str` to build a list
+of resources, each with a name from that list. When this is done, each resource
+will use the same set of parameters. The list of internal edges specified in the
+same resource block is created intelligently to have the appropriate edge for
+each separate resource.
+
+Using this construct is a veiled form of looping (iteration). This technique is
+one of many ways you can perform iterative tasks that you might have
+traditionally used a `for` loop for instead. This is preferred, because flow
+control is error-prone and can make for less readable code.
+
+##### Internal edges
+
 Resources may also declare edges internally. The edges may point to or from
 another resource, and may optionally include a notification. The four properties
 are: `Before`, `Depend`, `Notify` and `Listen`. The first two represent normal
@@ -197,7 +293,7 @@ send notifications. You may have multiples of these per resource, including
 multiple `Depend` lines if necessary. Each of these properties also supports the
 conditional inclusion `elvis` operator as well.
 
-For example, you may write is:
+For example, you may write:
 
 ```mcl
 $b = true # for example purposes
@@ -285,11 +381,12 @@ class baz($a str, $b) {
 Classes can also be nested within other classes. Here's a contrived example:
 
 ```mcl
+import "fmt"
 class c1($a, $b) {
 	# nested class definition
 	class c2($c) {
 		test $a {
-			stringptr => printf("%s is %d", $b, $c),
+			stringptr => fmt.printf("%s is %d", $b, $c),
 		}
 	}
 
@@ -305,6 +402,38 @@ time.
 Recursive classes are not currently supported and it is not clear if they will
 be in the future. Discussion about this topic is welcome on the mailing list.
 
+A class names can contain colons to indicate it is nested inside of the class in
+the same scope which is named with the prefix indicated by colon separation.
+Instead of needing to repeatedly indent the child classes, we can instead prefix
+them at the definition site (where created with the class keyword) with the name
+of the parent class, followed by a colon, to get the desired embedded sugar.
+
+For example, instead of writing:
+
+```mcl
+class base() {
+	class inner() {
+		class deepest() {
+		}
+	}
+}
+```
+
+You can instead write:
+
+```mcl
+class base() {
+}
+class base:inner() {
+}
+class base:inner:deepest() {
+}
+```
+
+Of course, you can only access any of the inner classes by first including
+(with the include keyword) a parent class, and then subsequently including the
+inner one.
+
 #### Include
 
 The `include` statement causes the previously defined class to produce the
@@ -317,6 +446,96 @@ parameters, then the same class can even be called with different signatures.
 Whether the output is useful and whether there is a unique type unification
 solution is dependent on your code.
 
+Classes can be included under a new scoped prefix by using the `as` field and an
+identifier. When used in this manner, the captured scope of the class at its
+definition site are made available in the scope of the include. Variables,
+functions, and child classes are all exported.
+
+Variables are available in the include scope:
+
+```mcl
+import "fmt"
+
+class c1 {
+	test "t1" {}	# gets pulled out
+	$x = "hello"	# gets exported
+}
+include c1 as i1
+
+test "print0" {
+	anotherstr => fmt.printf("%s", $i1.x),	# hello
+	onlyshow => ["AnotherStr",], # displays nicer
+}
+```
+
+Classes are also available in the new scope:
+
+```mcl
+import "fmt"
+class c1 {
+	test "t1" {}	# gets pulled out
+	$x = "hello"	# gets exported
+
+	class c0 {
+		test "t2" {}
+		$x = "goodbye"
+	}
+}
+include c1 as i1
+include i1.c0 as i0
+
+test "print0" {
+	anotherstr => fmt.printf("%s", $i1.x),	# hello
+	onlyshow => ["AnotherStr",], # displays nicer
+}
+test "print1" {
+	anotherstr => fmt.printf("%s", $i0.x),	# goodbye
+	onlyshow => ["AnotherStr",], # displays nicer
+}
+```
+
+Of course classes can be parameterized too, and those variables specified during
+the `include`.
+
+#### Import
+
+The `import` statement imports a scope into the specified namespace. A scope can
+contain variable, class, and function definitions. All are statements.
+Furthermore, since each of these have different logical uses, you could
+theoretically import a scope that contains an `int` variable named `foo`, a
+class named `foo`, and a function named `foo` as well. Keep in mind that
+variables can contain functions (they can have a type of function) and are
+commonly called lambdas.
+
+There are a few different kinds of imports. They differ by the string contents
+that you specify. Short single word, or multiple-word tokens separated by zero
+or more slashes are system imports. Eg: `math`, `fmt`, or even `math/trig`.
+Local imports are path imports that are relative to the current directory. They
+can either import a single `mcl` file, or an entire well-formed module. Eg:
+`file1.mcl` or `dir1/`. Lastly, you can have a remote import. This must be an
+absolute path to a well-formed module. The common transport is `git`, and it can
+be represented via an FQDN. Eg: `git://github.com/purpleidea/mgmt-example1/`.
+
+The namespace that any of these are imported into depends on how you use the
+import statement. By default, each kind of import will have a logic namespace
+identifier associated with it. System imports use the last token in their name.
+Eg: `fmt` would be imported as `fmt` and `math/trig` would be imported as
+`trig`. Local imports do the same, except the required `.mcl` extension, or
+trailing slash are removed. Eg: `foo/file1.mcl` would be imported as `file1` and
+`bar/baz/` would be imported as `baz`. Remote imports use some more complex
+rules. In general, well-named modules that contain a final directory name in the
+form: `mgmt-whatever/` will be named `whatever`. Otherwise, the last path token
+will be converted to lowercase and the dashes will be converted to underscores.
+The rules for remote imports might change, and should not be considered stable.
+
+In any of the import cases, you can change the namespace that you're imported
+into. Simply add the `as whatever` text at the end of the import, and `whatever`
+will be the name of the namespace. Please note that `whatever` is not surrounded
+by quotes, since it is an identifier, and not a `string`. If you'd like to add
+all of the import contents into the top-level scope, you can use the `as *` text
+to dump all of the contents in. This is generally not recommended, as it might
+cause a conflict with another identifier.
+
 ### Stages
 
 The mgmt compiler runs in a number of stages. In order of execution they are:
@@ -366,6 +585,10 @@ and can be used for other scenarios in which one statement or expression would
 be better represented by a larger AST. Most nodes in the AST simply return their
 own node address, and do not modify the AST.
 
+This stage also implements the class nesting when it finds class names with
+colons that should be nested inside of a base class. Currently this does modify
+the AST for efficiency and simplicity.
+
 #### Scope propagation
 
 Scope propagation passes the parent scope (starting with the top-level, built-in
@@ -375,6 +598,9 @@ without making any changes. The `ExprVar` node naturally consumes scope's and
 the `StmtProg` node cleverly passes the scope through in the order expected for
 the out-of-order bind logic to work.
 
+This step typically calls the ordering algorithm to determine the correct order
+of statements in a program.
+
 #### Type unification
 
 Each expression must have a known type. The unpleasant option is to force the
@@ -424,11 +650,11 @@ would like to propose a more logical or performant variant.
 
 #### Function graph generation
 
-At this point we have a fully type AST. The AST must now be transformed into a
+At this point we have a fully typed AST. The AST must now be transformed into a
 directed, acyclic graph (DAG) data structure that represents the flow of data as
 necessary for everything to be reactive. Note that this graph is *different*
 from the resource graph which is produced and sent to the engine. It is just a
-coincidence that both happen to be DAG's. (You don't freak out when you see a
+coincidence that both happen to be DAG's. (You aren't surprised when you see a
 list data structure show up in more than one place, do you?)
 
 To produce this graph, each node has a `Graph` method which it can call. This
@@ -436,9 +662,8 @@ starts at the top most node, and is called down through the AST. The edges in
 the graphs must represent the individual expression values which are passed
 from node to node. The names of the edges must match the function type argument
 names which are used in the definition of the corresponding function. These
-corresponding functions must exist for each expression node and are produced by
-calling that expression's `Func` method. These are usually called by the
-function engine during function creation and validation.
+corresponding functions must exist for each expression node and are produced as
+the vertices of this returned graph. This is built for the function engine.
 
 #### Function engine creation and validation
 
@@ -538,13 +763,13 @@ one value must be produced.
 
 ```golang
 Please see the example functions in
-[lang/funcs/core/](https://github.com/purpleidea/mgmt/tree/master/lang/funcs/core/).
+[lang/core/](https://github.com/purpleidea/mgmt/tree/master/lang/core/).
 ```
 
 ### Stream
 
 ```golang
-Stream() error
+Stream(context.Context) error
 ```
 
 Stream is called by the function engine when it is ready for your function to
@@ -553,29 +778,14 @@ value. Failure to produce at least one value will probably cause the function
 engine to hang waiting for your output. This function must close the `Output`
 channel when it has no more values to send. The engine will close the `Input`
 channel when it has no more values to send. This may or may not influence
-whether or not you close the `Output` channel.
+whether or not you close the `Output` channel. You must shutdown if the input
+context cancels.
 
 #### Example
 
 ```golang
 Please see the example functions in
-[lang/funcs/core/](https://github.com/purpleidea/mgmt/tree/master/lang/funcs/core/).
-```
-
-### Close
-
-```golang
-Close() error
-```
-
-Close asks the particular function to shutdown its `Stream()` function and
-return.
-
-#### Example
-
-```golang
-Please see the example functions in
-[lang/funcs/core/](https://github.com/purpleidea/mgmt/tree/master/lang/funcs/core/).
+[lang/core/](https://github.com/purpleidea/mgmt/tree/master/lang/core/).
 ```
 
 ### Polymorphic Function API

docs/on-the-web.md

@@ -44,3 +44,18 @@ if we missed something that you think is relevant!
 | James Shubin | blog | [Mgmt Configuration Language](https://purpleidea.com/blog/2018/02/05/mgmt-configuration-language/) |
 | James Shubin | video | [Recording from CfgMgmtCamp.eu 2018](https://www.youtube.com/watch?v=NxObmwZDyrI) |
 | Jonathan Gold | blog | [Go Netlink and Select](https://jonathangold.ca/blog/go-netlink-and-select/) |
+| James Shubin | video | [Recording from DevOpsDays Montreal 2018](https://www.youtube.com/watch?v=1i38c5cooHo) |
+| James Shubin | video | [Recording from FOSDEM Minimalistic Languages Devroom 2019](https://video.fosdem.org/2019/K.4.201/mgmtconfig.webm) |
+| James Shubin | video | [Recording from FOSDEM Infra Management Devroom 2019](https://video.fosdem.org/2019/UB2.252A/mgmt.webm) |
+| James Shubin | video | [Recording from FOSDEM Graph Processing Devroom 2019](https://video.fosdem.org/2019/H.1308/graph_mgmt_config.webm) |
+| James Shubin | video | [Recording from FOSDEM Virtualization Devroom 2019](https://video.fosdem.org/2019/H.2213/vai_real_time_virtualization_automation.webm) |
+| James Shubin | video | [Recording from FOSDEM Containers Devroom 2019](https://video.fosdem.org/2019/UA2.114/containers_mgmt.webm) |
+| James Shubin | video | [Recording from FOSDEM Monitoring Devroom 2019](https://video.fosdem.org/2019/UB2.252A/real_time_merging_of_config_management_and_monitoring.webm) |
+| James Shubin | blog | [Mgmt Configuration Language: Class and Include](https://purpleidea.com/blog/2019/07/26/class-and-include-in-mgmt/) |
+| James Shubin | video | [Recording from FOSDEM 2020, Main Track (History)](https://video.fosdem.org/2020/Janson/automation.webm) |
+| James Shubin | video | [Recording from FOSDEM 2020, Infra Management Devroom](https://video.fosdem.org/2020/UA2.120/mgmt.webm) |
+| James Shubin | video | [Recording from FOSDEM 2020, Minimalistic Languages Devroom](https://video.fosdem.org/2020/AW1.125/mgmtconfigmore.webm) |
+| James Shubin | video | [Recording from CfgMgmtCamp.eu 2020](https://www.youtube.com/watch?v=Kd7FAORFtsc) |
+| James Shubin | video | [Recording from CfgMgmtCamp.eu 2023](https://www.youtube.com/watch?v=FeRGRj8w0BU) |
+| James Shubin | video | [Recording from FOSDEM 2024, Golang Devroom](https://video.fosdem.org/2024/ud2218a/fosdem-2024-2575-single-binary-full-stack-provisioning.mp4) |
+| James Shubin | video | [Recording from CfgMgmtCamp.eu 2024](https://www.youtube.com/watch?v=vBt9lpGD4bc) |

docs/puppet-guide.md

@@ -1,166 +0,0 @@
-# Puppet guide
-
-`mgmt` can use Puppet as its source for the configuration graph.
-This document goes into detail on how this works, and lists
-some pitfalls and limitations.
-
-For basic instructions on how to use the Puppet support, see
-the [main documentation](documentation.md#puppet-support).
-
-## Prerequisites
-
-You need Puppet installed in your system. It is not important how you
-get it. On the most common Linux distributions, you can use packages
-from the OS maintainer, or upstream Puppet repositories. An alternative
-that will also work on OSX is the `puppet` Ruby gem. It also has the
-advantage that you can install any desired version in your home directory
-or any other location.
-
-Any release of Puppet's 3.x and 4.x series should be suitable for use with
-`mgmt`. Most importantly, make sure to install the `ffrank-mgmtgraph` Puppet
-module (referred to below as "the translator module").
-
-```
-puppet module install ffrank-mgmtgraph
-```
-
-Please note that the module is not required on your Puppet master (if you
-use a master/agent setup). It's needed on the machine that runs `mgmt`.
-You can install the module on the master anyway, so that it gets distributed
-to your agents through Puppet's `pluginsync` mechanism.
-
-### Testing the Puppet side
-
-The following command should run successfully and print a YAML hash on your
-terminal:
-
-```puppet
-puppet mgmtgraph print --code 'file { "/tmp/mgmt-test": ensure => present }'
-```
-
-You can use this CLI to test any manifests before handing them straight
-to `mgmt`.
-
-## Writing a suitable manifest
-
-### Unsupported attributes
-
-`mgmt` inherited its resource module from Puppet, so by and large, it's quite
-possible to express `mgmt` graphs in terms of Puppet manifests. However,
-there isn't (and likely never will be) full feature parity between the
-respective resource types. In consequence, a manifest can have semantics that
-cannot be transferred to `mgmt`.
-
-For example, at the time of writing this, the `file` type in `mgmt` had no
-notion of permissions (the file `mode`) yet. This lead to the following
-warning (among others that will be discussed below):
-
-```
-$ puppet mgmtgraph print --code 'file { "/tmp/foo": mode => "0600" }'
-Warning: cannot translate: File[/tmp/foo] { mode => "600" } (attribute is ignored)
-```
-
-This is a heads-up for the user, because the resulting `mgmt` graph will
-in fact not pass this information to the `/tmp/foo` file resource, and
-`mgmt` will ignore this file's permissions. Including such attributes in
-manifests that are written expressly for `mgmt` is not sensible and should
-be avoided.
-
-### Unsupported resources
-
-Puppet has a fairly large number of
-[built-in types](https://docs.puppet.com/puppet/latest/reference/type.html),
-and countless more are available through
-[modules](https://forge.puppet.com/). It's unlikely that all of them will
-eventually receive native counterparts in `mgmt`.
-
-When encountering an unknown resource, the translator module will replace
-it with an `exec` resource in its output. This resource will run the equivalent
-of a `puppet resource` command to make Puppet apply the original resource
-itself. This has quite abysmal performance, because processing such a
-resource requires the forking of at least one Puppet process (two if it
-is found to be out of sync). This comes with considerable overhead.
-On most systems, starting up any Puppet command takes several seconds.
-Compared to the split second that the actual work usually takes,
-this overhead can amount to several orders of magnitude.
-
-Avoid Puppet types that `mgmt` does not implement (yet).
-
-### Avoiding common warnings
-
-Many resource parameters in Puppet take default values. For the most part,
-the translator module just ignores them. However, there are cases in which
-Puppet will default to convenient behavior that `mgmt` cannot quite replicate.
-For example, translating a plain `file` resource will lead to a warning message:
-
-```
-$ puppet mgmtgraph print --code 'file { "/tmp/mgmt-test": }'
-Warning: File[/tmp/mgmt-test] uses the 'puppet' file bucket, which mgmt cannot do. There will be no backup copies!
-```
-
-The reason is that per default, Puppet assumes the following parameter value
-(among others)
-
-```puppet
-file { "/tmp/mgmt-test":
-	backup => 'puppet',
-}
-```
-
-To avoid this, specify the parameter explicitly:
-
-```bash
-puppet mgmtgraph print --code 'file { "/tmp/mgmt-test": backup => false }'
-```
-
-This is tedious in a more complex manifest. A good simplification is the
-following [resource default](https://docs.puppet.com/puppet/latest/reference/lang_defaults.html)
-anywhere on the top scope of your manifest:
-
-```puppet
-File { backup => false }
-```
-
-If you encounter similar warnings from other types and/or parameters,
-use the same approach to silence them if possible.
-
-## Configuring Puppet
-
-Since `mgmt` uses an actual Puppet CLI behind the scenes, you might
-need to tweak some of Puppet's runtime options in order to make it
-do what you want. Reasons for this could be among the following:
-
-* You use the `--puppet agent` variant and need to configure
-`servername`, `certname` and other master/agent-related options.
-* You don't want runtime information to end up in the `vardir`
-that is used by your regular `puppet agent`.
-* You install specific Puppet modules for `mgmt` in a non-standard
-location.
-
-`mgmt` exposes only one Puppet option in order to allow you to
-control all of them, through its `--puppet-conf` option. It allows
-you to specify which `puppet.conf` file should be used during
-translation.
-
-```
-mgmt run --puppet /opt/my-manifest.pp --puppet-conf /etc/mgmt/puppet.conf
-```
-
-Within this file, you can just specify any needed options in the
-`[main]` section:
-
-```
-[main]
-server=mgmt-master.example.net
-vardir=/var/lib/mgmt/puppet
-```
-
-## Caveats
-
-Please see the [README](https://github.com/ffrank/puppet-mgmtgraph/blob/master/README.md)
-of the translator module for the current state of supported and unsupported
-language features.
-
-You should probably make sure to always use the latest release of
-both `ffrank-mgmtgraph` and `ffrank-yamlresource` (the latter is
-getting pulled in as a dependency of the former).

docs/quick-start-guide.md

@@ -2,65 +2,123 @@
 
 ## Introduction
 
-This guide is intended for developers. Once `mgmt` is minimally viable, we'll
-publish a quick start guide for users too. If you're brand new to `mgmt`, it's
-probably a good idea to start by reading the
-[introductory article](https://purpleidea.com/blog/2016/01/18/next-generation-configuration-mgmt/)
-or to watch an [introductory video](https://www.youtube.com/watch?v=LkEtBVLfygE&html5=1).
-Once you're familiar with the general idea, please start hacking...
+This guide is intended for users and developers. If you're brand new to `mgmt`,
+it's probably a good idea to start by reading an
+[introductory article about the engine](https://purpleidea.com/blog/2016/01/18/next-generation-configuration-mgmt/)
+and an [introductory article about the language](https://purpleidea.com/blog/2018/02/05/mgmt-configuration-language/).
+[There are other articles and videos available](on-the-web.md) if you'd like to
+learn more or prefer different formats. Once you're familiar with the general
+idea, or if you prefer a hands-on approach, please start hacking...
 
-## Quick start
+## Getting mgmt
 
-### Installing golang
+You can either build `mgmt` from source, or you can download a pre-built
+release. There are also some distro repositories available, but they may not be
+up to date. A pre-built release is the fastest option if there's one that's
+available for your platform. If you are developing or testing a new patch to
+`mgmt`, or there is not a release available for your platform, then you'll have
+to build your own.
 
-* You need golang version 1.9 or greater installed.
+### Downloading a pre-built release:
+
+This method is not recommended because those packages are now very old.
+
+The latest releases can be found [here](https://github.com/purpleidea/mgmt/releases/).
+An alternate mirror is available [here](https://dl.fedoraproject.org/pub/alt/purpleidea/mgmt/releases/).
+
+Make sure to verify the signatures of all packages before you use them. The
+signing key can be downloaded from [https://purpleidea.com/contact/#pgp-key](https://purpleidea.com/contact/#pgp-key)
+to verify the release.
+
+If you've decided to install a pre-build release, you can skip to the
+[Running mgmt](#running-mgmt) section below!
+
+### Building a release:
+
+You'll need some dependencies, including `golang`, and some associated tools.
+
+#### Installing golang
+
+* You need golang version 1.20 or greater installed.
 	* To install on rpm style systems: `sudo dnf install golang`
 	* To install on apt style systems: `sudo apt install golang`
 	* To install on macOS systems install [Homebrew](https://brew.sh)
 	and run: `brew install go`
 * You can run `go version` to check the golang version.
-* If your distro is tool old, you may need to [download](https://golang.org/dl/)
+* If your distro is too old, you may need to [download](https://golang.org/dl/)
 a newer golang version.
 
-### Setting up golang
+#### Setting up golang
 
-* If you do not have a GOPATH yet, create one and export it:
+* You can skip this step, as your installation will default to using `~/go/`,
+but if you do not have a `GOPATH` yet and want one in a custom location, create
+one and export it:
 
-```
+```shell
 mkdir $HOME/gopath
 export GOPATH=$HOME/gopath
 ```
 
 * You might also want to add the GOPATH to your `~/.bashrc` or `~/.profile`.
-* For more information you can read the [GOPATH documentation](https://golang.org/cmd/go/#hdr-GOPATH_environment_variable).
+* For more information you can read the
+[GOPATH documentation](https://golang.org/cmd/go/#hdr-GOPATH_environment_variable).
 
-### Getting the mgmt code and dependencies
+#### Getting the mgmt code and associated dependencies
 
-* Download the `mgmt` code into the GOPATH, and switch to that directory:
+* Download the `mgmt` code and switch to that directory:
 
-```
-mkdir -p $GOPATH/src/github.com/purpleidea/
-cd $GOPATH/src/github.com/purpleidea/
-git clone --recursive https://github.com/purpleidea/mgmt/
-cd $GOPATH/src/github.com/purpleidea/mgmt
+```shell
+git clone --recursive https://github.com/purpleidea/mgmt/ ~/mgmt/
+cd ~/mgmt/
 ```
 
-* Add $GOPATH/bin to $PATH
+* Add `$GOPATH/bin` to `$PATH`
 
-```
+```shell
 export PATH=$PATH:$GOPATH/bin
 ```
 
 * Run `make deps` to install system and golang dependencies. Take a look at
-`misc/make-deps.sh` for details.
-* Run `make build` to get a freshly built `mgmt` binary.
+`misc/make-deps.sh` if you want to see the details of what it does.
 
-### Running mgmt
+#### Building mgmt
 
-* Run `time ./mgmt run --lang examples/lang/hello0.mcl --tmp-prefix` to try out
-a very simple example!
+* Now run `make` to get a freshly built `mgmt` binary. If this succeeds, you can
+proceed to the [Running mgmt](#running-mgmt) section below!
+
+### Installing a distro release
+
+Installation of `mgmt` from distribution packages currently needs improvement.
+They are not always up-to-date with git master and as such are not recommended.
+At the moment we have:
+* [COPR](https://copr.fedoraproject.org/coprs/purpleidea/mgmt/) (currently dead)
+* [Arch](https://aur.archlinux.org/packages/mgmt/) (currently stale)
+
+Please contribute more and help improve these! We'd especially like to see a
+Debian package!
+
+### Building from a container:
+
+This method avoids polluting your workstation with the dependencies for the
+build. Here is an example using Fedora, Podman and Buildah:
+
+```shell
+git clone --recursive https://github.com/purpleidea/mgmt/ ~/mgmt/
+cd ~/mgmt/docker
+buildah build -f Dockerfile-fedora.build -t mgmt_build
+podman run -d -it --name mgmt_build localhost/mgmt_build
+podman cp mgmt_build:/src/github.com/purpleidea/mgmt/mgmt /tmp/mgmt
+sudo mv /tmp/mgmt /usr/local/bin  # be sure this is in your $PATH
+sudo chown root:root /usr/local/bin/mgmt
+```
+
+## Running mgmt
+
+* Run `mgmt run --tmp-prefix lang examples/lang/hello0.mcl` to try out a very
+simple example! If you built it from source, you'll need to use `./mgmt` from
+the project directory.
 * Look in that example file that you ran to see if you can figure out what it
-did!
+did! You can press `^C` to exit `mgmt`.
 * Have fun hacking on our future technology and get involved to shape the
 project!
 
@@ -68,118 +126,3 @@ project!
 
 Please look in the [examples/lang/](../examples/lang/) folder for some more
 examples!
-
-## Vagrant
-
-If you would like to avoid doing the above steps manually, we have prepared a
-[Vagrant](https://www.vagrantup.com/) environment for your convenience. From the
-project directory, run a `vagrant up`, and then a `vagrant status`. From there,
-you can `vagrant ssh` into the `mgmt` machine. The MOTD will explain the rest.
-
-## Using Docker
-
-Alternatively, you can check out the [docker-guide](docs/docker-guide.md) in
-order to develop or deploy using docker.
-
-## Information about dependencies
-
-Software projects have a few different kinds of dependencies. There are _build_
-dependencies, _runtime_ dependencies, and additionally, a few extra dependencies
-required for running the _test_ suite.
-
-### Build
-
-* `golang` 1.9 or higher (required, available in some distros and distributed
-as a binary officially by [golang.org](https://golang.org/dl/))
-
-### Runtime
-
-A relatively modern GNU/Linux system should be able to run `mgmt` without any
-problems. Since `mgmt` runs as a single statically compiled binary, all of the
-library dependencies are included. It is expected, that certain advanced
-resources require host specific facilities to work. These requirements are
-listed below:
-
-| Resource | Dependency        | Version                     | Check version with                                        |
-|----------|-------------------|-----------------------------|-----------------------------------------------------------|
-| augeas   | augeas-devel      | `augeas 1.6` or greater     | `dnf info augeas-devel` or `apt-cache show libaugeas-dev` |
-| file     | inotify           | `Linux 2.6.27` or greater   | `uname -a`                                                |
-| hostname | systemd-hostnamed | `systemd 25` or greater     | `systemctl --version`                                     |
-| nspawn   | systemd-nspawn    | `systemd ???` or greater    | `systemctl --version`                                     |
-| pkg      | packagekitd       | `packagekit 1.x` or greater | `pkcon --version`                                         |
-| svc      | systemd           | `systemd ???` or greater    | `systemctl --version`                                     |
-| virt     | libvirt-devel     | `libvirt 1.2.0` or greater  | `dnf info libvirt-devel` or `apt-cache show libvirt-dev`  |
-| virt     | libvirtd          | `libvirt 1.2.0` or greater  | `libvirtd --version`                                      |
-
-For building a visual representation of the graph, `graphviz` is required.
-
-To build `mgmt` without augeas support please run:
-`GOTAGS='noaugeas' make build`
-
-To build `mgmt` without libvirt support please run:
-`GOTAGS='novirt' make build`
-
-To build `mgmt` without docker support please run:
-`GOTAGS='nodocker' make build`
-
-To build `mgmt` without augeas, libvirt or docker support please run:
-`GOTAGS='noaugeas novirt nodocker' make build`
-
-## Binary Package Installation
-
-Installation of `mgmt` from distribution packages currently needs improvement.
-They are not always up-to-date with git master and as such are not recommended.
-At the moment we have:
-* [COPR](https://copr.fedoraproject.org/coprs/purpleidea/mgmt/)
-* [Arch](https://aur.archlinux.org/packages/mgmt/)
-
-Please contribute more! We'd especially like to see a Debian package!
-
-## OSX/macOS/Darwin development
-
-Developing and running `mgmt` on macOS is currently not supported (but not
-discouraged either). Meaning it might work but in the case it doesn't you would
-have to provide your own patches to fix problems (the project maintainer and
-community are glad to assist where needed).
-
-There are currently some issues that make `mgmt` less suitable to run for provisioning
-macOS. But as a client to provision remote servers it should run fine.
-
-Since the primary supported systems are Linux and these are the environments
-tested for it is wise to run these suites during macOS development as well. To
-ease this Docker can be leveraged ([Docker for Mac](https://docs.docker.com/docker-for-mac/)).
-
-Before running any of the commands below create the development Docker image:
-
-```
-docker/scripts/build-development
-```
-
-This image requires updating every time dependencies (`make-deps.sh`) change.
-
-Then to run the test suite:
-
-```
-docker run --rm -ti \
-	-v $PWD:/go/src/github.com/purpleidea/mgmt/ \
-	-w /go/src/github.com/purpleidea/mgmt/ \
-	purpleidea/mgmt:development \
-	make test
-```
-
-For convenience this command is wrapped in `docker/scripts/exec-development`.
-
-Basically any command can be executed this way. Because the repository source is
-mounted into the Docker container invocation will be quick and allow rapid
-testing, example:
-
-```
-docker/scripts/exec-development test/test-shell.sh load0.sh
-```
-
-Other examples:
-
-```
-docker/scripts/exec-development make build
-docker/scripts/exec-development ./mgmt run --tmp-prefix --lang examples/lang/load0.mcl
-```

docs/release-notes/0.0.10

@@ -0,0 +1,51 @@
+I've just released version 0.0.10 of mgmt!
+
+NEWS
+
+57 files changed, 1991 insertions(+), 752 deletions(-)
+
+* There's a new resource called `KV`. Short examples exist, but I
+haven't yet published a whole integration showing the usefulness.
+
+* A major race was fixed. The issue of what to do with BackPokes during
+start/pause was never previously solved. I had this as an open issue on
+my whiteboard for a while, and I finally got some time to work through
+it. The answer wasn't that difficult, but I think it was shrouded in
+some tunnel vision. See the commit messages and source comments for the
+details.
+
+* The GAPI grew four new associated World API methods: StrWatch,
+StrGet, StrSet, StrDel, and the associated etcd backed implementations.
+These are quite useful when combined with the KV resource.
+
+* There are now P/V style counting semaphores available as metaparams.
+This is particularly cool because the implementation is (AFAIK,
+assuming no bugs) dead-lock free! This is mentioned in my recent blog
+post.
+
+* See the git log for more NEWS, and sorry for anything notable I left
+out!
+
+BUGS
+
+* There's a `concurrent map write` bug in the semaphore implementation
+which is fixed in git master. Since it was a race, it was only caught
+after this release was made. I should also figure out if the sema check
+should go after the BackPoke or not.
+
+MISC
+
+We're still looking for new contributors, and there are easy, medium
+and hard issues available! You're also welcome to suggest your own!
+Please join us in #mgmtconfig on Freenode IRC, or ping this list if
+you'd like help getting started!
+
+THANKS
+
+Thanks (alphabetically) to everyone who contributed to the latest
+release:
+James Shubin, Julien Pivotto, Michael Borden.
+We had 3 unique committers since 0.0.9, and have had 30 overall.
+
+Happy hacking,
+James

docs/release-notes/0.0.11

@@ -0,0 +1,59 @@
+I've just released version 0.0.11 of mgmt!
+
+NEWS
+
+ 20 files changed, 579 insertions(+), 126 deletions(-)
+
+* Added a missing mutex around the semaphore map which prevents
+occasional panics
+
+* Removed exec pollint param which is not needed because of the poll
+metaparam
+
+* Fixed a state rechecking bug in exec resource (things are faster now)
+
+* Fixed the major annoyance of exec resources receiving main's signals.
+If we would ^C the main mgmt, the child processes would receive this
+too which was incorrect.
+
+* Fixed the deadlock on resource errors. This meant that previously if
+a resource failed, we would deadlock the graph from shutting down. This
+was bad and I'm glad it's now fixed. Sorry about that!
+
+* Improved the backpoke logic to not require semaphores since we used
+to take the lock even when we were going to backpoke which was
+unnecessary.
+
+* Added fast pausing to the graph. This means that a ^C or a pause
+transition used to wait for the whole graph to roll through, but it now
+finishes after the currently running resources finish executing. Read
+the commit messages for more background here including the discussion
+about a possible Interrupt() addition to the resource API.
+
+* Prometheus support has been updated!
+
+* See the git log for more NEWS, and sorry for anything notable I left
+out!
+
+BUGS
+
+* We're in pretty good shape now. There are some small issues, but
+nothing major that I don't know about.
+
+MISC
+
+We're still looking for new contributors, and there are easy, medium
+and hard issues available! You're also welcome to suggest your own!
+Please join us in #mgmtconfig on Freenode IRC, or ping this list if
+you'd like help getting started!
+
+THANKS
+
+Thanks (alphabetically) to everyone who contributed to the latest
+release:
+James Shubin, Julien Pivotto
+We had 2 unique committers since 0.0.10, and have had 30 overall.
+
+Happy hacking,
+James
+@purpleidea

docs/release-notes/0.0.12

@@ -0,0 +1,87 @@
+I've just released version 0.0.12 of mgmt!
+
+Sorry if I've been more quiet than usual, I've had to focus a lot of my
+time on GlusterFS related features ($dayjob official directives) and
+work there.
+
+One goal is to try and use libmgmt to either wholly or partially re-
+implement glusterd. As a result, a lot of my upstream focus has been
+re-prioritized to features needed for that effort.
+
+I wrote a PoC called gd3: https://github.com/purpleidea/gd3
+(It has now bit-rotted compared to upstream mgmt, but is easy to fix.)
+The initial scope looks like it will be much smaller, but hopefully
+this is interesting to you too. Ping me if you'd like to help.
+
+We desperately need your contributions if we're going to get mgmt
+standalone into a MVP. To motivate you, there's some great new stuff
+that landed since 0.0.11, including:
+
+* a great new YAML parser from contributor Mildred
+
+* a huge pgraph refactor (to remove internal deps and cycles)
+
+* a great amount of new tests and testing
+
+and so much more...
+
+NEWS
+
+ 76 files changed, 7549 insertions(+), 4269 deletions(-)
+
+* The svc resource now also supports user services
+
+* There's a fabulous new yaml parser that fixes a longstanding issue in
+my original code. Try it with --yaml2. I'll remove the old one
+eventually. Thanks to contributor Mildred for this great work!
+
+* Refactored the lib/ etcd usage into the GAPI's for a cleaner main
+
+* World API grew some new methods for libmgmt users
+
+* pgraph refactor and cleanup, now it's a fairly clean standalone pkg
+
+* pgraph functions to flatten/merge in subgraphs along with examples
+
+* Giant resource refactor to hopefully make things more logical, and to
+simplify the resource API. This also introduces the mgraph struct to
+add the higher level graph knowledge outside of pgraph.
+
+* A partial implementation of a "Graph" (recursive subgraph?) resource!
+See the code for details, as help is wanted to finish this. This should
+help elucidate what the most elegant design for the mgmt core should
+be.
+
+* Send/Recv support for the exec resource as output, stdout, and stderr
+
+* GAPI improvements to support exit requests and fast pausing
+
+* AutoEdge API improvements including a fix+test for a regression
+
+* A possible fix for the possible etcd server startup race
+
+* A fun amount of new tests all over including for gometalinter
+
+* See the git log for more NEWS, and for anything notable I left out!
+
+BUGS
+
+* We fixed a bunch of stuff, and added more tests!
+
+MISC
+
+We're still looking for new contributors, and there are easy, medium
+and hard issues available! You're also welcome to suggest your own!
+Please join us in #mgmtconfig on Freenode IRC, or ping this list if
+you'd like help getting started!
+
+THANKS
+
+Thanks (alphabetically) to everyone who contributed to the latest
+release:
+James Shubin, Julien Pivotto, Mildred Ki'Lya
+We had 3 unique committers since 0.0.11, and have had 30 overall.
+
+Happy hacking,
+James
+@purpleidea

docs/release-notes/0.0.13

@@ -0,0 +1,89 @@
+I've just released version 0.0.13 of mgmt!
+
+I guess this is an appropriate number for a scary October release.
+
+We recently re-licensed to the more permissive GPL. I put a lot of
+thought into it, and also wrote up a short blog post about some of the
+reasoning. It's here:
+
+https://ttboj.wordpress.com/2017/10/17/copyleft-is-dead-long-live-copyl
+eft/
+
+If you read to the end you'll also see that Red Hat is not funding my
+mgmt work anymore. This is too bad, and means I'll only have small
+amounts of personal time available to work on this. If you'd like to
+help fund my work, or know someone who'd like to, please let me know!
+
+Having said that, there's some great new stuff that landed since
+0.0.12, including:
+
+* new resources from new contributor Jonathan Gold (aws, group, user)
+
+* an HCL frontend from new contributor Chris McKenzie
+
+* polish in a number of places including in the nspawn resource
+
+and so much more... If you'd prefer to have releases more often, then
+please let me know! Lastly, the language and a number of associated
+parts are on its way. I hope to push this monster patch to git master
+before February.
+
+It's also worth mentioning that we have 17 resources now! wow.
+
+NEWS
+
+ 140 files changed, 3921 insertions(+), 848 deletions(-)
+
+* Many improvements to tests, testing and small fixes to avoid false-
+failures on travis.
+
+* golint now reports 0 issues.
+
+* An HCL frontend if you'd prefer that to the YAML. Also a great
+example of how to plug in a new frontend.
+
+* An update to golang 1.8 as the minimum version required.
+
+* Bump the etcd version to 3.2.6+ -- Looking forward to a 3.3 release
+which should probably include some patches we upstreamed.
+
+* Addition of new user and group resources. These also include a bunch
+of automatic edges.
+
+* Addition of an AWS resource! I've wanted this for a while, as it
+demonstrates nicely how event based cloud resources can fit nicely into
+our design. There's still a lot to do, and we have some suggestions for
+Amazon too. If you have a contact there, please put me in touch!
+
+* Our nspawn resource is more polished now.
+
+* See the git log for more NEWS, and for anything notable I left out!
+
+MISC
+
+We're still looking for new contributors, and there are easy, medium
+and hard issues available! You're also welcome to suggest your own!
+Please join us in #mgmtconfig on Freenode IRC, or ping this list if
+you'd like help getting started!
+
+MENTORING
+
+We offer mentoring for new golang/mgmt hackers who want to get
+involved. This is free and friendly. You get to improve your skills,
+and we get some patches in return. Ping me off-list for details.
+
+THANKS
+
+Thanks (alphabetically) to everyone who contributed to the latest
+release:
+AdnanLFC, Arthur Mello, ChrisMcKenzie, Dennis Kliban, Ismael Puerto,
+James Shubin, Jonathan Gold, Juan Luis de Sousa-Valadas Castaño, Juan-
+Luis de Sousa-Valadas Castaño (although I suspect the last two are the
+same ;))
+
+We had 9 unique committers since 0.0.12, and have had 38 overall.
+Run 'git log 0.0.12..0.0.13' to see what has changed since 0.0.12
+
+Happy hacking,
+James
+@purpleidea

docs/release-notes/0.0.14

@@ -0,0 +1,81 @@
+I've just released version 0.0.14 of mgmt!
+
+> 118 files changed, 2688 insertions(+), 974 deletions(-)
+
+There's some great new stuff that landed since
+0.0.13, including:
+
+* amazon AWS EC2 resource is now in git master.
+
+* more automatic edges from new contributor Guillaume Herail (xiu)
+
+* a move to golang 1.8 or higher
+
+and so much more... This will probably be the last release before the
+language lands in git master. It's a pretty giant patch coming :/
+
+NEWS
+
+* We're > 1k stars on GitHub now. It's a silly metric, but ¯\_(ツ)_/¯
+
+* Jonathan Gold has done a lot of hard work on the AWS EC2 resource,
+and it's now in git master. There are still many things we'd like to
+do, but it's a great start on what is a *monster* resource, and
+hopefully it will inspire others to get involved.
+
+In particular, it was a great learning experience (for me in
+particular!) about how bad the EC2 golang API is. There are some
+notable design bugs we found, and if anyone from Amazon engineering
+would like to reach out to us, we'd love to provide some constructive
+ideas.
+
+* Guillaume Herail (xiu) wrote some really nice patches, and picked up
+on the autoedges API very quickly. Hopefully he'll have time to work on
+even more!
+
+* Paul Morgan sent us some nice shell fixups-- many more exist in an
+open PR that didn't make it into this release. Hopefully we'll get
+those merged by 0.0.15
+
+* Felix Frank did a few Puppet compiler fix ups. I think he's been
+refreshing his work with new resources recently...
+
+* We're looking for help writing Google, DigitalOcean, Rackspace, etc
+resources if anyone is interested, reach out to us. Particularly if
+there is support from those organizations as well.
+
+* See the git log for more NEWS, and for anything notable I left out!
+
+EVENTS
+
+There are a bunch of upcoming mgmt talks and events! Stay tuned for
+details in the coming email, but TL;DR: Linux Conf Australia, FOSDEM, &
+CfgMgmtCamp.eu -- from three different speakers, and including a
+hackathon too!
+
+MISC
+
+We're still looking for new contributors, and there are easy, medium
+and hard issues available! You're also welcome to suggest your own!
+Please join us in #mgmtconfig on Freenode IRC, or ping this list if
+you'd like help getting started! Expect many new tagged #mgmtlove
+issues within the month.
+
+MENTORING
+
+We offer mentoring for new golang/mgmt hackers who want to get
+involved. This is free and friendly. You get to improve your skills,
+and we get some patches in return. Ping me off-list for details.
+
+THANKS
+
+Thanks (alphabetically) to everyone who contributed to the latest
+release:
+Felix Frank, Guillaume Herail, James Shubin, Jonathan Gold,
+jonathangold, Julien Pivotto, Paul Morgan, Toshaan Bharvani
+We had 8 unique committers since 0.0.13, and have had 41 overall.
+run 'git log 0.0.13..0.0.14' to see what has changed since 0.0.13
+
+Happy hacking,
+James
+@purpleidea

docs/release-notes/0.0.15

@@ -0,0 +1,116 @@
+I've just released version 0.0.15 of mgmt!
+
+> 328 files changed, 29869 insertions(+), 943 deletions(-)
+
+(Yeah, that's almost 30k+ LOC)
+
+There's some great new stuff that landed since 0.0.14, including:
+
+* THE LANGUAGE (mcl)
+
+* "Deploys": a distributed way to push code into your cluster
+
+* Scheduling (as a reactive function)
+
+* Better testing
+
+* a move to etcd 3.3+ and golang 1.9+
+
+and so much more... This is a monster release. Please try out the
+language and all the other new features today :)
+
+NEWS
+
+* We released the language. Please play around with it :) It's time to
+get used to this cool new paradigm. Learn more from the...
+
+Blog post:
+https://purpleidea.com/blog/2018/02/05/mgmt-configuration-language/
+
+Video:
+https://www.youtube.com/watch?v=NxObmwZDyrI
+
+Docs:
+https://github.com/purpleidea/mgmt/blob/master/docs/language-guide.md
+
+Function guide:
+https://github.com/purpleidea/mgmt/blob/master/docs/function-guide.md
+
+And tons of code all over git master. Check the lang/ folder.
+
+* There is a reactive scheduler in the language. Use your imagination,
+or play around with:
+https://github.com/purpleidea/mgmt/blob/3ad7097c8aa7eab7f895aab9af22338
+c0cf82986/lang/funcs/core/schedule_polyfunc.go#L18
+
+* There is a "deploys" feature. It's not documented yet. You should
+poke around if you're curious. Consider this an early soft release.
+
+* There is a FS implementation to store files in a POSIX-like layer on
+top of etcd. It's used by deploys. It needs more tests though :)
+
+* The language grew two "simple" API's for implementing functions, so
+that new functionality can be exposed in the mgmt language.
+
+* The language grew two ways to specify edges between resources: either
+internal to the resource, or externally as standalone edges.
+
+* The language now supports optional resource parameters via the
+"elvis" operator. This keeps things type safe and avoids needing an
+"undef" or "nil" in the language. This operator also works for edge
+declarations.
+
+* New contributor Johan Bloemberg has been on fire sending patches!
+He has already made some great improvements to our Makefile for
+testing, and the addition of the env* functions in the language, with
+much more code pending in open PR's.
+
+* Initial debian packaging has been added. It now needs a maintainer to
+build, upload, and love it :)
+
+* We have an early emacs major mode for "mcl", our language.
+
+* Lots of new documentation has been added. Particularly for developers
+wanting to contribute to the project.
+
+* We're looking for help writing Google, DigitalOcean, Rackspace, etc
+resources if anyone is interested, reach out to us. Particularly if
+there is support from those organizations as well.
+
+* See the git log for more NEWS, and for anything notable I left out!
+
+MISC
+
+We're still looking for new contributors, and there are easy, medium
+and hard issues available! You're also welcome to suggest your own!
+Please join us in #mgmtconfig on Freenode IRC, or ping this list if
+you'd like help getting started! For details please see:
+
+https://github.com/purpleidea/mgmt/blob/master/docs/faq.md#how-do-i-con
+tribute-to-the-project-if-i-dont-know-golang
+
+Many new tagged #mgmtlove issues were tagged:
+https://github.com/purpleidea/mgmt/issues?q=is%3Aissue+is%3Aopen+label%
+3Amgmtlove
+
+MENTORING
+
+We offer mentoring for new golang/mgmt hackers who want to get
+involved.
+This is free and friendly. You get to improve your skills,
+and we get
+some patches in return. Ping me off-list for details.
+
+THANKS
+
+Thanks (alphabetically) to everyone who contributed to the latest
+release:
+Carsten Thiel, dsx, James Shubin, Joe Julian, Johan Bloemberg, Jonathan
+Gold, jonathangold, karimb, Oliver Frommel, Peter Oliver, Toshaan
+Bharvani, Wim
+We had 12 unique committers since 0.0.14, and have had 48 overall.
+run 'git log 0.0.14..0.0.15' to see what has changed since 0.0.14
+
+Happy hacking,
+James
+@purpleidea

docs/release-notes/0.0.16

@@ -0,0 +1,104 @@
+I've just released version 0.0.16 of mgmt!
+
+> 220 files changed, 14243 insertions(+), 9491 deletions(-)
+
+Woo...
+
+There's some great new stuff that landed since 0.0.15, including:
+
+* A giant engine re-write! (Makes resource writing more elegant too.)
+
+* New resources!!
+
+* New language features!!!
+
+and so much more... This is a monster release. Please try out the new
+features today :)
+
+NEWS
+
+* New resources include: net, mount and docker:container. Jonathan was
+responsible for all of these. Please take them for a spin! He's looking
+for a job too, and would probably be happy to get paid to work on mgmt.
+
+* We're > 1.5k stars on GitHub now. It's a silly metric, but ¯\_(ツ)_/¯
+
+* A giant engine refactoring/re-write was done. This cleaned up the
+code significantly, and made it more elegant to write resources.
+Unfortunately there is one small bug that I missed and that I haven't
+fixed yet. It rarely happens except during some of our tests during
+shutdown, which causes intermittent failures. It shouldn't block you
+playing with mgmt.
+
+* The language "class" and "include" statements have been added. These
+are important pieces for writing reusable modules which are coming
+soon. Try them out! (This comes with a bunch of tests too.)
+
+* We have an integration testing framework. It's pretty cool, it spins
+up a full mgmt cluster and runs stuff. Try it out or add some tests.
+
+* I had fun fixing a big bug: 06ee05026b0c743d19c7d62675f8ddeabdc8dd4f
+
+* I removed the remote execution functionality from core. I realized it
+could be re-written as a resource, and it was also in the way from some
+other cleanups that were more important. Half the new code is done,
+ping me if this is a priority for you or you want to help.
+
+* I also removed the HCL front-end, because mcl is usable enough to be
+more fun to play with, and I wanted to refactor some code. If someone
+really wants it back, let me know.
+
+* We have some release building scripts in git master, so you can now
+download pre-built (with fpm) RPM, DEB, or PACMAN packages! They're
+signed too. https://github.com/purpleidea/mgmt/releases/tag/0.0.16
+
+* We're looking for help writing Google, DigitalOcean, Rackspace, etc
+resources if anyone is interested, reach out to us. Particularly if
+there is support from those organizations as well.
+
+* Many other bug fixes, changes, etc...
+
+* See the git log for more NEWS, and for anything notable I left out!
+
+MISC
+
+I took a bit of a break recently to catch up on some life stuff, but I
+think I'm back on track. While git master hasn't been especially busy,
+there's an active feature branch at feat/import which contains some fun
+stuff, with a very WIP giant patch still sitting on my machine. I hope
+to finish it up soon and then do another release. That branch contains
+one of the last big features before I'll really be ready to run mgmt on
+my personal servers!
+
+We're still looking for new contributors, and there are easy, medium
+and hard issues available! You're also welcome to suggest your own!
+Please join us in #mgmtconfig on Freenode IRC, or ping this list if
+you'd like help getting started! For details please see:
+
+https://github.com/purpleidea/mgmt/blob/master/docs/faq.md#how-do-i-con
+tribute-to-the-project-if-i-dont-know-golang
+
+Many new tagged #mgmtlove issues were tagged:
+https://github.com/purpleidea/mgmt/issues?q=is%3Aissue+is%3Aopen+label%
+3Amgmtlove
+
+MENTORING
+
+We offer mentoring for new golang/mgmt hackers who want to get
+involved.
+This is free and friendly. You get to improve your skills,
+and we get
+some patches in return. Ping me off-list for details.
+
+THANKS
+
+Thanks (alphabetically) to everyone who contributed to the latest
+release:
+Alan Jenkins, James Shubin, jesus m. rodriguez, Jonathan Gold,
+jonathangold, Lauri Ojansivu, phaer
+We had 7 unique committers since 0.0.15, and have had 52 overall.
+run 'git log 0.0.15..0.0.16' to see what has changed since 0.0.15
+
+Happy hacking,
+James
+@purpleidea

docs/release-notes/0.0.17

@@ -0,0 +1,111 @@
+I've just released version 0.0.17 of mgmt!
+
+> 269 files changed, 13281 insertions(+), 1633 deletions(-)
+
+There's some very useful stuff that landed since 0.0.16, including:
+
+* Modules and import system now exists!
+* A lot of tests, fixes and a huge new test infra was added
+* Merging puppet with mcl code is now possible (madness!)
+* We made a small change to the cli UI
+* Bump to golang 1.10 (please update your environments)
+
+And much more...
+
+DOWNLOAD
+
+Prebuilt binaries are available here:
+https://github.com/purpleidea/mgmt/releases/tag/0.0.17
+
+NEWS
+
+* One of the biggest missing features was the lack of a module/import
+system. After some initial exploration on what turned out to be a dead-
+end, I found what I think is a very elegant approach, which is now in
+this release. Please try it out, there are docs available. I hope to
+write a blog post about it soon.
+
+(There's one additional "kind" of import that I'm considering, similar
+to a macro "#include", that I might add. To be determined. Let me know
+if you find anything missing as of today.)
+
+* Felix added the first version of his mcl+puppet frontend. This should
+allow you to more cleverly merge legacy puppet environments with mcl
+code. It's fantastic, take a look.
+
+* Jonathan added a systemd-timer resource. This is a great replacement
+for cron.
+
+* We changed the CLI ui to improve the determinism of the frontend
+selection. Basically we changed from: `mgmt run --lang code.mcl` to:
+`mgmt run lang --lang code.mcl`. Remember to put --tmp-prefix after
+`run` directly where it is used.
+
+* We made a whole bunch of cleanups to the test infra, added new test
+infra for testing complex mcl modules and the import/module system, and
+of course added new tests.
+
+* You can pass a list of strings as the resource name to build that
+many resources. (Looping/iteration!)
+
+* You can specify all the metaparams and auto-* properties in mcl now.
+
+* Native mcl code can be used to write imported core code.
+
+* There was a bug that snuck into the pkg res. This has now been fixed.
+
+* A small, long-time copy+pasta error bug was fixed in Exec.
+
+* Virtually all the imports of the "log" package are at the top-level
+now. This will make moving to a new logger easier in the future. I have
+an innovative logger idea that I have a design for that I'll eventually
+get to.
+
+* A few workarounds for occasional test failures were added. Some
+legacy code needs a cleanup, and it's not done yet. Fortunately, none
+of these issues seem to occur in real-life as far as I can tell, and
+are caused by closing down mgmt at weird times.
+
+* Found a bug (now fixed) in the upstream lexer Yikes! See:
+57ce3fa587897d74634c1216af67dd42252c64e5
+
+* We're looking for help writing Amazon, Google, DigitalOcean, etc
+resources if anyone is interested, reach out to us. Particularly if
+there is support from those organizations as well.
+
+* Many other bug fixes, changes, etc...
+
+* See the git log for more NEWS, and for anything notable I left out!
+
+MISC
+
+We're still looking for new contributors, and there are easy, medium
+and hard issues available! You're also welcome to suggest your own!
+Please join us in #mgmtconfig on Freenode IRC, or ping this list if
+you'd like help getting started! For details please see:
+
+https://github.com/purpleidea/mgmt/blob/master/docs/faq.md#how-do-i-con
+tribute-to-the-project-if-i-dont-know-golang
+
+Many new tagged #mgmtlove issues were tagged:
+https://github.com/purpleidea/mgmt/issues?q=is%3Aissue+is%3Aopen+label%
+3Amgmtlove
+
+MENTORING
+
+We offer mentoring for new golang/mgmt hackers who want to get
+involved. This is fun and friendly! You get to improve your skills,
+and we get some patches in return. Ping me off-list for details.
+
+THANKS
+
+Thanks (alphabetically) to everyone who contributed to the latest
+release:
+Felix Frank, James Shubin, Jonathan Gold, Kevin Kuehler, Michael Lesko-
+Krleza, Tom Payne, Vincent Membré
+We had 7 unique committers since 0.0.16, and have had 56 overall.
+run 'git log 0.0.16..0.0.17' to see what has changed since 0.0.16
+
+Happy hacking,
+James
+@purpleidea

docs/release-notes/0.0.18

@@ -0,0 +1,106 @@
+I've just released version 0.0.18 of mgmt!
+
+> 202 files changed, 5606 insertions(+), 1880 deletions(-)
+
+There's some great stuff that landed since 0.0.17, including:
+
+* A re-write of the core engine algorithm
+* Tests are very stable
+* At least three old bugs were killed
+* An infra to write tests for individual resources was added
+
+And much more...
+
+This adds a significant amount of polish and bug fixes to mgmt. We're
+on the home stretch to MVP!
+
+DOWNLOAD
+
+Prebuilt binaries are available here:
+https://github.com/purpleidea/mgmt/releases/tag/0.0.18
+
+NEWS
+
+* There was a rare race that would panic the engine on shutdown. This
+only seemed to happen in CPU/system call starved environments like
+travis-ci. This was due to some ignorance when writing this early part
+of the code base. The algorithm has been re-written, and this also
+removed the use of at least one mutex. Things should be stable now, and
+also much faster. (Although perf was never an issue.)
+
+* A hidden race/deadlock in the exec resource was found and killed.
+Woo! Some new tools to help find these and other problems are in misc/
+
+* The early converger code was re-written. I was not as knowledgeable
+about golang in the very beginning, and this code needed refreshing. It
+contained a rare deadlock which needed to be killed.
+
+* Toshaan added an uptime() function.
+
+* Julien added a method for generating some simple functions for the
+language.
+
+* Lander added two new functions.
+
+* James added a new readfile() function, and other examples.
+
+* The template function now allows you to use imported functions. They
+use underscores instead of periods for namespace separation due to a
+limitation in the template library.
+
+* Kevin and I killed a tricky race in the SocketSet code! Woo :) Kevin
+also added a great cpucount() fact!
+
+* James gave a number of presentations at FOSDEM. Some recordings are
+available: https://purpleidea.com/talks/
+
+* We're looking for help writing Amazon, Google, DigitalOcean, etc
+resources if anyone is interested, reach out to us. Particularly if
+there is support from those organizations as well.
+
+* Many other bug fixes, changes, etc...
+
+* See the git log for more NEWS, and for anything notable I left out!
+
+BUGS
+
+* An unfortunate bug in the type unification code was found. This can
+cause small code bases to take a lot of ram/cpu to run. This will be
+prioritized in an upcoming release. Until then you'll have to avoid
+fancy type unification. (Specify types you know when it has to guess.)
+
+If efficient type unification algorithms are your specialty, please let
+us know, we'd like your help!
+
+MISC
+
+We're still looking for new contributors, and there are easy, medium
+and hard issues available! You're also welcome to suggest your own!
+Please join us in #mgmtconfig on Freenode IRC, or ping this list if
+you'd like help getting started! For details please see:
+
+https://github.com/purpleidea/mgmt/blob/master/docs/faq.md#how-do-i-con
+tribute-to-the-project-if-i-dont-know-golang
+
+Many tagged #mgmtlove issues exist:
+https://github.com/purpleidea/mgmt/issues?q=is%3Aissue+is%3Aopen+label%
+3Amgmtlove
+
+MENTORING
+
+We offer mentoring for new golang/mgmt hackers who want to get
+involved. This is fun and friendly! You get to improve your skills,
+and we get some patches in return. Ping me off-list for details.
+
+THANKS
+
+Thanks (alphabetically) to everyone who contributed to the latest
+release:
+Felix Frank, James Shubin, Jeff Waugh, Johan Bloemberg, Julien Pivotto,
+Kevin Kuehler, Lander Van den Bulcke, Toshaan Bharvani
+We had 8 unique committers since 0.0.17, and have had 58 overall.
+run 'git log 0.0.17..0.0.18' to see what has changed since 0.0.17
+
+Happy hacking,
+James
+@purpleidea

docs/release-notes/0.0.19

@@ -0,0 +1,131 @@
+I've just released version 0.0.19 of mgmt!
+
+> 361 files changed, 10451 insertions(+), 3919 deletions(-)
+
+This is a very important (and huge) release and has some important
+fixes that landed since 0.0.18, including:
+
+* A huge re-write of the elastic etcd clustering code base
+* A significant improvement in the type unification algorithm
+* An important import/class scoping bug was discovered and fixed
+* New mcl functions and resource improvements
+
+And much more...
+
+What comes next is just polish, new features and small bug fixes!
+
+DOWNLOAD
+
+Prebuilt binaries are available here:
+https://github.com/purpleidea/mgmt/releases/tag/0.0.19
+
+NEWS
+
+* A giant etcd re-write was completed and merged. The elastic
+clustering algorithm is not perfect, however it should suffice for most
+use cases, and it's always possible to point mgmt at an external etcd
+cluster if you don't understand the limitations of the automatic
+clustering algorithm. The important part is that the core code is much
+cleaner now, so hopefully races and bugs of ignorance are gone now. :)
+
+* I found an unfortunate bug in the type unification algorithm that
+severely impacted performance for some types of code bases. This is now
+fixed, and I hope we should not experience problems again! Special
+thanks to Sam for talking me through the problem and understanding the
+space better! Woo \o/
+
+* An important import/class scoping bug was fixed. Thanks to Nicolas
+for the bug report. We also added tests for this too!
+
+* Nicolas also added our first os detection function. os.is_debian and
+os.is_redhat are now in core. Get your favourite os added today!
+
+* The polymorphic len function can also check str length.
+
+* The exec resource got a big cleanup. It also learned the interrupt
+trait so that long running commands can be forcefully killed if need
+be.
+
+* A fancy new test infra for testing functions over time was added.
+Anytime we want to check our individual FRP functions are working as
+expected, this is an easy way to add a test. This way, if we ever find
+a bug, we can drop in a test with the fix. This actually helped find a
+very subtle bug in readfile that nobody had experienced yet!
+
+* File res with state => exists but no content now performs as
+expected.
+
+* Improved send/recv, since it was neglected a bit. Hopefully it ends
+up being a useful primitive.
+
+* Added a new synchronization primitive that I'm called
+SubscribedSignal. I found it very useful for building some of my code,
+and I hope you'll find it useful too. I'd offer it upstream to the sync
+package if Google didn't force their crappy CLA nonsense on everyone
+who wanted to send a patch. :/ Death by 1000 paper cuts, I guess.
+
+* Added a match function in the new regexp core package. Try it out and
+add some more functions!
+
+* Wouter has been testing mgmt and filling all sorts of useful bug
+reports. We fixed at least one related to a report, and more are
+planned for the next release. Wouter also sent in one cleanup patch to
+remove some dead code. Welcome to the project!
+
+* We're looking for help writing Amazon, Google, DigitalOcean, etc
+resources if anyone is interested, reach out to us. Particularly if
+there is support from those organizations as well.
+
+* Many other bug fixes, changes, etc...
+
+* See the git log for more NEWS, and for anything notable I left out!
+
+BUGS
+
+* There are a few known issues, in particular with some over eager
+checking done in the Validate portion of two resources, that should
+actually be runtime checks in CheckApply. As a result, if you intend to
+change some state during the graph execution, the resource won't see
+it. These should be very easy to fix if someone is interested in
+writing the patch!
+
+TALKS
+
+* James will be presenting at this year's OSDC in Berlin. There will be
+a workshop: https://osdc.de/events/mgmt-config-workshop/
+and a talk:
+https://osdc.de/events/mgmt-config-the-future-of-your-autonomous-datacentre/
+Sign up soon if you want to guarantee a spot, as they're limited!
+
+MISC
+
+We're still looking for new contributors, and there are easy, medium
+and hard issues available! You're also welcome to suggest your own!
+Please join us in #mgmtconfig on Freenode IRC, or ping this list if
+you'd like help getting started! For details please see:
+
+https://github.com/purpleidea/mgmt/blob/master/docs/faq.md#how-do-i-con
+tribute-to-the-project-if-i-dont-know-golang
+
+Many tagged #mgmtlove issues exist:
+https://github.com/purpleidea/mgmt/issues?q=is%3Aissue+is%3Aopen+label%
+3Amgmtlove
+
+MENTORING
+
+We offer mentoring for new golang/mgmt hackers who want to get
+involved. This is fun and friendly! You get to improve your skills,
+and we get some patches in return. Ping me off-list for details.
+
+THANKS
+
+Thanks (alphabetically) to everyone who contributed to the latest
+release:
+Adam Sigal, Felix Frank, James Shubin, Jonathan Gold, Michael Schubert,
+Mitch Fossen, Nicolas Charles, Wouter Dullaert
+We had 8 unique committers since 0.0.18, and have had 63 overall.
+run 'git log 0.0.18..0.0.19' to see what has changed since 0.0.18
+
+Happy hacking,
+James
+@purpleidea

docs/release-notes/0.0.20

@@ -0,0 +1,145 @@
+I've just released version 0.0.20 of mgmt!
+
+> 295 files changed, 8585 insertions(+), 1413 deletions(-)
+
+This was a very challenging release but it includes many useful changes
+since 0.0.19, including:
+
+* Function values / lambdas exist and are first-class
+* Over 70 new tests have been added!
+* Core functions, classes, and globals can now be written in mcl code
+* A new compiler "Ordering" step was added
+
+And much more...
+
+DOWNLOAD
+
+Prebuilt binaries are available here:
+https://github.com/purpleidea/mgmt/releases/tag/0.0.20
+
+NEWS
+
+* After a long and challenging road, I finally got function values and
+lambdas merged. If you find any bugs, please let me know! You can now
+treat functions as first-class values, and even use them as closures by
+capturing variable state.
+
+* Over 70 new tests were added, most of which test the behaviour of the
+new functions.
+
+* Core packages always allowed you to write new functions in pure
+golang, but now you can implement new functions, classes, and even
+variables in native mcl code! It's still desirable and perhaps
+efficient in some cases to want native golang implementations, but all
+good programs self-host some of their stdlib in their own language
+eventually.
+
+* A new compiler step called "Ordering" was added. It's hidden inside
+the SetScope step, but now lets us generate code ordering graphs and
+determine exactly what to run first.
+
+* Light copying of Node's allows more correct and efficient function
+graphs that can share common vertices and edges. For example, if two
+different closures capture a variable $x, they'll both use the same
+copy when running the function, since the compiler can prove if they're
+identical.
+
+* Improved the type system slightly to allow advanced type comparisons.
+
+* The type unification algorithm was improved. Hopefully it should
+solve all scenarios without needing the recursive solver which was very
+slow. If you find a case that isn't speedy, please let us know!
+
+* Added subtest listing by using -short and -v in a test. This lets you
+know what's available and how to run individual sub tests easily.
+
+* Support for the systemd STATE_DIRECTORY and xdg cache dir was added
+by new contributor John! Thanks!
+
+* New contributor Adam added a pgraph test.
+
+* A bug in the systemd mount resource was fixed. Hopefully it should
+work correctly now.
+
+* A bug that prevented us from allowing nested system imports was
+fixed. Nest away! This will pave the way for us to automatically import
+most of the golang standard library by doing: import "golang/regexp" or
+golang/whatever".
+
+* Added an example showing that unicode is allowed in strings.
+
+* Fixed a rare race in the engine.
+
+* Added some new core functions including math.mod and datetime
+improvements.
+
+* Changed the API to remove the use of --lang. This avoids the
+stuttering.
+
+* Moved to golang 1.11 and etcd 3.3.13. The later includes a fix for an
+un-catchable error scenario which we fixed in etcd.
+
+* Improved the pgraph library significantly so that we can generate
+better graphs with accurate vertices based on the vertex pointers.
+
+* Added ArchLinux OS family detection.
+
+* We're looking for help writing Amazon, Google, DigitalOcean, etc
+resources if anyone is interested, reach out to us. Particularly if
+there is support from those organizations as well.
+
+* Many other bug fixes, changes, etc...
+
+* See the git log for more NEWS, and for anything notable I left out!
+
+BUGS
+
+* There are a few known issues with some disabled code paths in the new
+function value code. These issues don't *need* to be fixed, but if they
+are, then we should see a slight performance increase. Happy to have
+someone dig into these, and they shouldn't bother anyone.
+
+* Some of the pre-built binaries might not work on your system. We need
+to start building them with the right dependencies so that `ldd`
+related things are happy. For now, please try building yourself if the
+build doesn't work for you, or help improve our build system.
+
+TALKS
+
+I'll be giving a talk at an upcoming mini-conference in Montreal. If
+you're interested in attending, please let me know.
+If you'd like to give an mgmt talk somewhere, please let me know!
+
+MISC
+
+We're still looking for new contributors, and there are easy, medium
+and hard issues available! You're also welcome to suggest your own!
+Please join us in #mgmtconfig on Freenode IRC, or ping this list if
+you'd like help getting started! For details please see:
+
+https://github.com/purpleidea/mgmt/blob/master/docs/faq.md#how-do-i-con
+tribute-to-the-project-if-i-dont-know-golang
+
+Many tagged #mgmtlove issues exist:
+https://github.com/purpleidea/mgmt/issues?q=is%3Aissue+is%3Aopen+label%
+3Amgmtlove
+
+MENTORING
+
+We offer mentoring for new golang/mgmt hackers who want to get
+involved. This is fun and friendly! You get to improve your skills,
+and we get some patches in return. Ping me off-list for details.
+
+THANKS
+
+Thanks (alphabetically) to everyone who contributed to the latest
+release:
+Adam Sigal, Christian Rebischke, Felix Frank, James Shubin, Jan
+Martens, Johan Bloemberg, John Hooks, Ward Vandewege
+We had 8 unique committers since 0.0.19, and have had 67 overall.
+run 'git log 0.0.19..0.0.20' to see what has changed since 0.0.19
+
+
+Happy hacking,
+James
+@purpleidea

docs/release-notes/0.0.21

@@ -0,0 +1,133 @@
+I've just released version 0.0.21 of mgmt!
+
+> 178 files changed, 4351 insertions(+), 829 deletions(-)
+
+This was a very lonely release but it includes some very useful
+additions since 0.0.20, including:
+
+* The first mgmt meme!
+* Working distro packages for Fedora, Debian, Ubuntu and Arch!
+* Reversible resources!
+* A deploy package to let you read files from the active deploy
+* Improved file resource behaviours
+
+And much more...
+
+DOWNLOAD
+
+Prebuilt binaries are available here:
+https://github.com/purpleidea/mgmt/releases/tag/0.0.21
+
+NEWS
+
+* Someone made a cool mgmt meme. Look in art/mgmt_*_meme.jpg
+
+* Distro packages are now built properly, so they should all work now.
+We have builds for Fedora-30, Fedora-29, Debian-10, Ubuntu-Bionic, and
+Archlinux. If you'd like a build for a different distro/version please
+let me know.
+
+* We finally got rid of the old `Compare(...) bool` API, and moved to
+`Cmp(...) error`. We'll now get more useful information from Res
+compares when they differ. Thanks to new contributor Donald Bakong for
+working on this. He's ramping up his golang contributions, so we expect
+to see more from him in the future!
+
+* We now have reversible resources. Basically if you create a resource
+and specify the reverse metaparam, eg: `Meta:reverse => true`, then if
+a resource is removed (either because a new version of code doesn't
+have it anymore OR more importantly if it was inside an `if` block that
+became false) then the engine will perform some "reverse" action for
+it. For a file, if it was added, we'll remove it. If it was edited,
+we'll undo the edit. If we added ugo+w, we'll remove that. And so on.
+The engine bits are done, and as well so have the file resource bits.
+It should be easy to add this for any other resource where it makes
+sense. This will likely be a very powerful feature that we use a lot.
+
+* The file resource was changed slightly so that by default the "state"
+is undefined. As a result, if you want a file to be created and none is
+present, you need to specify the state. Otherwise specifying "content"
+will only edit a file if it already exists, and otherwise be an error.
+It turns out this is actually a better behaviour, even if it's not
+necessarily intuitive for puppet users. It turns out it simplifies the
+code drastically and makes the reversible file resource much more
+logical. It seems that Puppet and Ansible got this wrong, but Cfengine
+got it right. IIRC. Do you agree? (Look at the code!)
+
+* We now catch CR \r characters in code so that you don't wonder why
+the compiler is telling you about unexpected whitespace. This should
+make your life easier.
+
+* You can now read files from within the deploy. This can be used for
+templates or anything else. This was one of the last missing things
+that was blocking me from writing useful mcl modules.
+
+* Fixed a copy-pasta bug where the != operator (for strings only) was
+actually doing an ==. Woops! The good news is that we've been shaking
+out silly bugs because I've been using mgmt more and more. Hopefully
+there aren't any woops ones like this left!
+
+* A bunch of function, class, and include tests were added. We're
+getting really well tested!
+
+* We're looking for help writing Amazon, Google, DigitalOcean, etc
+resources if anyone is interested, reach out to us. Particularly if
+there is support from those organizations as well.
+
+* Many other bug fixes, changes, etc...
+
+* See the git log for more NEWS, and for anything notable I left out!
+
+BUGS
+
+* Function values as args don't work yet. This is blocking us from
+implementing functions like map/reduce/filter, because they'd want to
+receive a function as input. TBH, I'm a bit tunnel-visioned on this,
+because I'm not the compilers genius that you are. If you can help,
+please let me know. I'll be posting a bunch of test cases that show
+what's needed shortly.
+
+* Three patches have been submitted to mkosi to support the image
+building I've been doing. They're not merged yet, so you'll have to
+apply them yourself if you want to make your own distro images. This
+isn't a major requirement anyone should have, but if they're not
+merged, we'll store them here and apply them as needed.
+
+TALKS
+
+I'll be in Belgium for FOSDEM and CfgMgmtCamp in 2020. If I'm lucky
+I'll have at least one mgmt presentation. I might also consider going
+to DevConf.cz if I get a talk accepted. Feel to ping me if you'd like
+to hack, get consulting, training, etc while I'm in Europe!
+
+MISC
+
+We're still looking for new contributors, and there are easy, medium
+and hard issues available! You're also welcome to suggest your own!
+Please join us in #mgmtconfig on Freenode IRC, or ping this list if
+you'd like help getting started! For details please see:
+
+https://github.com/purpleidea/mgmt/blob/master/docs/faq.md#how-do-i-con
+tribute-to-the-project-if-i-dont-know-golang
+
+Many tagged #mgmtlove issues exist:
+https://github.com/purpleidea/mgmt/issues?q=is%3Aissue+is%3Aopen+label%
+3Amgmtlove
+
+MENTORING
+
+We offer mentoring for new golang/mgmt hackers who want to get
+involved. This is fun and friendly! You get to improve your skills,
+and we get some patches in return. Ping me off-list for details.
+
+THANKS
+
+Thanks (alphabetically) to everyone who contributed to the latest
+release:
+bjanssens, Donald Bakong, James Shubin
+We had 3 unique committers since 0.0.20, and have had 69 overall.
+run 'git log 0.0.20..0.0.21' to see what has changed since 0.0.20
+
+Happy hacking,
+James
+@purpleidea

docs/release-notes/0.0.22

@@ -0,0 +1,160 @@
+I've just released version 0.0.22 of mgmt!
+
+> 579 files changed, 17984 insertions(+), 3136 deletions(-)
+
+Compared to the last release, this is a monster. Previously:
+
+> 178 files changed, 4351 insertions(+), 829 deletions(-)
+
+I apologize for not doing a release earlier, but to be quite honest,
+I've been busy, the people who are playing with mgmt are doing their
+own builds anyways, and there are still some unimplemented, pre-
+production features missing.
+
+Also, I started a new job. News on my blog and how (if at all) it
+affects mgmt is there.
+
+This is still an incredibly important release, so let's cover some of
+the points!
+
+* This is the last release before we switch to go.mod
+* IRC channel moved to #mgmtconfig on libera.chat after Freenode died
+* New resources including tftp, dhcp, and http (all as servers, wow!)
+* New string interpolation implementation with many tests
+* Resource fields can accept complex structs and other types now
+* Improved type unification solver and new invariants like generator
+* A new polymorphic function API interface
+
+And much more...
+
+DOWNLOAD
+
+Prebuilt binaries are NOT available here for this release:
+https://github.com/purpleidea/mgmt/releases/tag/0.0.22
+
+NEWS
+
+* The file resource supports building files from other "fragments".
+This is magic and automatic and real-time. Docs and examples are in the
+repo.
+
+* The file resource (and others) have a new trait and queryable API to
+make decisions based on what other resources are in the graph.
+
+* The file resource has a "purge" option to remove unmanaged files from
+a managed directory.
+
+* A lot of built-in functions are auto-generated from the stdlib. Most
+things you would want are now present, particularly if they're pure
+functions.
+
+* There's a new consul KV resource.
+
+* File resources support symbolic modes now!
+
+* New tftp related resources are now present. They're great and I use
+them to provision things!
+
+* A docker image resource was added.
+
+* First-class constants now exist. So you can do
+$const.res.file.state.exists instead of typing "exists" which is prone
+to typos. This is more verbose, but it's safer if that's your priority.
+
+* We found a bug with fuzzing! Cool, thanks Patrick!
+
+* We have dhcp server related resources. This is pretty cool when
+combined with the tftp resource and you can provision a lot of stuff
+from your laptop and one binary now.
+
+* We also have http server resources. Combined with the tftp and dhcp
+resources mgmt starts to look like a powerful tool to greenfield a new
+datacentre and then take over and manage it continuous. All from a
+single, type-safe, code base. Of course you can do other things with
+this, and I'm looking forward to seeing the ideas that I haven't
+thought of yet! PS: An http:ui has been partially implemented too. Ping
+if you want to know more.
+
+* Resource fields couldn't previously accept anonymous structs as types
+because of how golang built its reflect library. Joe found an elegant
+workaround, thanks!
+
+* The type unification solver was improved to support some new
+invariants. This makes a lot of new things possible, and was done to
+support new complex functions including the eventual addition of map,
+reduce, and filter. One of the new invariants is a "generator"
+invariant, so that unification can take into account the entire
+relevant parts of the AST. It's not a textbook CS implementation, but
+it's based on sound theory I think, and it seems to work great. If you
+find an edge case, please let us know.
+
+* The polymorphic function interface was changed to use the new
+unification logic. This makes a lot more sense. All the functions have
+been ported to the new interface.
+
+* 42
+
+* We're looking for help writing Amazon, Google, DigitalOcean, etc
+resources if anyone is interested, reach out to us. Particularly if
+there is support from those organizations as well.
+
+* Many other bug fixes, changes, etc...
+
+* See the git log for more NEWS, and for anything notable I left out!
+
+BUGS
+
+* Function values as args don't work yet. This is blocking us from
+implementing functions like map/reduce/filter, because they'd want to
+receive a function as input. TBH, I'm a bit tunnel-visioned on this,
+because I'm not the compilers genius that you are. If you can help,
+please let me know. I'll be posting a bunch of test cases that show
+what's needed shortly.
+
+(This is the exact message I wrote last time. I've made a lot of
+progress since then, but motivation here has been low. Reach out if you
+can help.)
+
+TALKS
+
+Hopefully CfgMgmtCamp in 2022 will be back on. If we're lucky and safe,
+maybe I can travel there. TBD... Feel to ping me if you'd like to hack,
+talk, whatever if I'm in Europe.
+
+MISC
+
+We're still looking for new contributors, and there are easy, medium
+and hard issues available! You're also welcome to suggest your own!
+Please join us in #mgmtconfig on Libera IRC, or ping this list if
+you'd like help getting started! For details please see:
+
+https://github.com/purpleidea/mgmt/blob/master/docs/faq.md#how-do-i-con
+tribute-to-the-project-if-i-dont-know-golang
+
+Many tagged #mgmtlove issues exist:
+https://github.com/purpleidea/mgmt/issues?q=is%3Aissue+is%3Aopen+label%
+3Amgmtlove
+
+Although asking in IRC is the best way to find something to work on.
+
+MENTORING
+
+We offer mentoring for new golang/mgmt hackers who want to get
+involved. This is fun and friendly! You get to improve your skills,
+and we get some patches in return. Ping me off-list for details.
+
+THANKS
+
+Thanks (alphabetically) to everyone who contributed to the latest
+release:
+Adam Sigal, Ahmed Al-Hulaibi, David Randall, Derek Buckley, Donald
+Bakong, Felix Frank, Francois Rompre-Lanctot, Ivan Pejić, James Shubin,
+Jean-Philippe Evrard, Jimmy Tang, Joe Groocock, Jonathan Gold, Julien
+Pivotto, Kenneth Hoste, Matthew Lesko-Krleza, Patrick Meyer, viq, Yohan
+Belval
+We had 19 unique committers since 0.0.21, and have had 82 overall.
+run 'git log 0.0.21..0.0.22' to see what has changed since 0.0.21
+
+Happy hacking,
+James
+@purpleidea

docs/release-notes/0.0.23

@@ -0,0 +1,102 @@
+I've just released version 0.0.23 of mgmt!
+
+>  424 files changed, 7051 insertions(+), 2256 deletions(-)
+
+This is a fairly quiet release, and I'm mostly doing this to have a
+permanent tag before I start really breaking git master.
+
+I'd like to apologize for things being kind of quiet. I've had to focus
+on life and making money to pay my bills, and I've been struggling a
+bit to complete some of the tougher algorithmic parts that I think are
+necessary for a solid MVP. Hopefully I will succeed, but to do so, it's
+going to be easier if I break git master and then sort things out
+later.
+
+I'm feeling optimistic about the future, although help with some
+complex concurrent programming would certainly be appreciated.
+
+With that, here are a few highlights from the release:
+
+* We're using go.mod, but I'm not keeping it up-to-date regularly yet.
+
+* We have an os.system(`cmd`) function!
+
+* We replaced go-bindata with the embed package.
+
+* We added a hetzner vm resource.
+
+* We did a giant lang/ package refactor.
+
+* Printf supports %v now (any type).
+
+And much more...
+
+DOWNLOAD
+
+Prebuilt binaries are NOT available here for this release:
+https://github.com/purpleidea/mgmt/releases/tag/0.0.23
+
+NEWS
+
+* We have our Libera IRC channel bridged to Matrix, so you can join
+that way too.
+
+* We're looking for help writing Amazon, Google, DigitalOcean, Hetzner,
+etc, resources if anyone is interested, reach out to us. Particularly
+if there is support from those organizations as well.
+
+* Many other bug fixes, changes, etc...
+
+* See the git log for more NEWS, and for anything notable I left out!
+
+BUGS
+
+* Function values as args don't work yet. This is blocking us from
+implementing functions like map/reduce/filter, because they'd want to
+receive a function as input.
+
+(This is the exact message I wrote last time. I've made a lot of
+progress since then, but motivation here has been low. Reach out if you
+can help.)
+
+TALKS
+
+I don't have anything planned until CfgMgmtCamp 2024. If you'd like to
+book me for a private event, or sponsor my travel for your conference,
+please let me know.
+
+MISC
+
+We're still looking for new contributors, and there are easy, medium
+and hard issues available! You're also welcome to suggest your own!
+Please join us in #mgmtconfig on Libera IRC or Matrix, or ping this
+list if you'd like help getting started! For details please see:
+
+https://github.com/purpleidea/mgmt/blob/master/docs/faq.md#how-do-i-con
+tribute-to-the-project-if-i-dont-know-golang
+
+Many tagged #mgmtlove issues exist:
+https://github.com/purpleidea/mgmt/issues?q=is%3Aissue+is%3Aopen+label%
+3Amgmtlove
+
+Although asking in IRC is the best way to find something to work on.
+
+MENTORING
+
+We offer mentoring for new golang/mgmt hackers who want to get
+involved. This is fun and friendly! You get to improve your skills,
+and we get some patches in return. Ping me off-list for details.
+
+THANKS
+
+Thanks (alphabetically) to everyone who contributed to the latest
+release:
+dantefromhell, James Shubin, Jef Masereel, Joe Groocock, Samuel
+Gélineau
+We had 5 unique committers since 0.0.22, and have had 85 overall.
+run 'git log 0.0.22..0.0.23' to see what has changed since 0.0.22
+
+
+Happy hacking,
+James
+@purpleidea

docs/release-notes/0.0.24

@@ -0,0 +1,161 @@
+I've just released version 0.0.24 of mgmt!
+
+
+> 600 files changed, 13622 insertions(+), 6907 deletions(-)
+
+This is a huge and hugely important release! It has been a long time
+coming. We have lambdas! I could not have done this without the
+unrelentingly supportive and brilliant Samuel Gélineau.
+
+Getting past the blockers and tricky code changes was all thanks to
+him. Cleanups, polish and making it more golang idiomatic was my doing.
+(The easy stuff.) If Sam wanted to be a golang expert, he could have
+done it all, but it was more sensible than I do all the mundane and
+filler stuff.
+
+With that, here are a few highlights from the release:
+
+* We have working lambdas, including iter.map =D
+
+* We have a new function engine!
+
+* We have improved type unification!
+
+* We have an improved resource engine!
+
+* We have improved many of the internal API's.
+
+* We have so many tests.
+
+And much more...
+
+DOWNLOAD
+
+Prebuilt binaries are NOT available here for this release:
+https://github.com/purpleidea/mgmt/releases/tag/0.0.24
+
+NEWS
+
+* Our main tests are now in the excellent .txtar format. This makes
+things much easier to manage.
+
+* The iter.map function can be named as such due to parser tricks! No
+need to name it xmap anymore!
+
+* The unification solver has been greatly improved. We can infer a lot
+more about function types.
+
+* The resources API uses the context package for closing Watch and
+returning early from CheckApply. The next step would be to remove all
+the resource-specific timeout code and make that a metaparam.
+
+* A new sync primitive has been added in case you'd like to use it
+somewhere. We'd love help adding an even more complex one. Look in
+util/sync.go for more information.
+
+* Sam added some beautiful type inference debugging that makes things
+easier for those familiar with the standard literature.
+
+* An important bug in standalone etcd has been fixed. While embedded
+etcd and automatic clustering isn't "supported" (it's buggy) the
+status-quo of using your own etcd cluster is stable, and you can even
+use the embedded etcd server in standalone mode...
+
+* This means you can run `mgmt etcd` and get the standard etcd binary
+behviour that you'd get from running `etcd` normally. This makes it
+easy to use both together since you only need to transport one binary
+around. (And maybe mgmt will do that for you!)
+
+* I fixed and cleaned up some sketchy code in the resource engine. I
+had been unmotivated to fix this for a while because I really wanted
+lambdas first, but now that they are in, I took a good look at the
+code, made some fixes, and I'm really happy with it now.
+
+* Metaparams are appropriately stateful between graph switches now.
+Retry is the easy example. Limit/Burst need to be ported if you care
+about these fine details.
+
+* A RetryReset metaparam has been added. This is another good example
+of how powerful metaparams are and how much potential there is for
+building systems with these and future ones.
+
+* A bunch of internal API's have been updated. This makes it better for
+function and resource writers! Some GAPI changes also got pushed
+through that make things clearer for those reading code.
+
+* We have a listlookup function. It's still missing syntactic sugar
+though!
+
+* Our new function graph engine is called "dage". I think it's pretty
+clever. There's a chance there is still a bug inside, but it's unclear.
+Please report any issues. If you have some large machines I can test
+very large and fast graphs on, please let me know.
+
+* Lambdas really work now! The txn and ref/gc code is pretty fantastic.
+
+* Many bugs have been killed!
+
+* We're looking for help writing Amazon, Google, DigitalOcean, Hetzner,
+etc, resources if anyone is interested, reach out to us. Particularly
+if there is support from those organizations as well.
+
+* Many other bug fixes, changes, etc...
+
+* See the git log for more NEWS, and for anything notable I left out!
+
+BUGS/TODO
+
+* Function values getting _passed_ to resources doesn't work yet, but
+it's not a blocker, but it would definitely be useful. We're looking
+into it.
+
+* Function graphs are unnecessarily dynamic. We might make them more
+static so that we don't need as many transactions. This is really a
+compiler optimization and not a bug, but it's something important we'd
+like to have.
+
+* Running two Txn's during the same pause would be really helpful. I'm
+not sure how much of a performance improvement we'd get from this, but
+it would sure be interesting to build. If you want to build a fancy
+synchronization primitive, then let us know! Again this is not a bug.
+
+TALKS
+
+I don't have anything planned until CfgMgmtCamp 2024. If you'd like to
+book me for a private event, or sponsor my travel for your conference,
+please let me know.
+
+MISC
+
+We're still looking for new contributors, and there are easy, medium
+and hard issues available! You're also welcome to suggest your own!
+Please join us in #mgmtconfig on Libera IRC or Matrix, or ping this
+list if you'd like help getting started! For details please see:
+
+https://github.com/purpleidea/mgmt/blob/master/docs/faq.md#how-do-i-con
+tribute-to-the-project-if-i-dont-know-golang
+
+Many tagged #mgmtlove issues exist:
+https://github.com/purpleidea/mgmt/issues?q=is%3Aissue+is%3Aopen+label%
+3Amgmtlove
+
+Although asking in IRC/matrix is the best way to find something to work
+on.
+
+MENTORING
+
+We offer mentoring for new golang/mgmt hackers who want to get
+involved. This is fun and friendly! You get to improve your skills,
+and we get some patches in return. Ping me off-list for details.
+
+THANKS
+
+Thanks (alphabetically) to everyone who contributed to the latest
+release:
+James Shubin, Kaushal, Laurent Indermuehle, Ofek Atar, Samuel Gélineau
+We had 5 unique committers since 0.0.23, and have had 88 overall.
+run 'git log 0.0.23..0.0.24' to see what has changed since 0.0.23
+
+Happy hacking,
+James
+@purpleidea

docs/release-notes/0.0.9

@@ -0,0 +1,71 @@
+I've just released 0.0.9!
+
+Release tags are now signed with my GPG key.
+
+From now on, I'm going to try and write a release email to the list
+with information about each release. Here we go!
+
+NEWS
+
+* There were far more commits that went into this release than I was
+expecting. I delayed a while because of some nasty races and deadlocks
+I encountered when cleaning some cruft out of the engine.
+
+134 files changed, 5394 insertions(+), 2168 deletions(-)
+
+* We grew initial integration with Prometheus thanks to new contributor
+`roidelapluie`.
+
+* The file resource can now chown/chmod files! Thanks to new
+contributor `mildred`.
+
+* The virt resource can now hotplug/hotunplug cpus. This made for some
+dope demos at CfgMgmtCamp. Special thanks to pkrempa for helping me
+with the libvirt API.
+
+* Native testing of all the golang code is now enabled again. I broke
+this when I split mgmt up into multiple packages. (Sorry!) Golang tests
+in foo_test.go should now run and be tested automatically. I'd
+especially like it if someone wrote some for pgraph/pgraph.go:
+GraphSync().
+
+* Documentation has been significantly improved. We have a resource
+guide (for writing new native resources) and much more. We also now
+build the docs as a RTD project:
+https://mgmt.readthedocs.io/en/master/
+Patches welcome to make it style as pretty as GitHub markdown does.
+
+* See the git log for more NEWS, and sorry for anything notable I left
+out!
+
+BUGS
+
+* I hope to resolve this soon, but it shouldn't block further
+development on git master for now. The issue: if you make extremely
+high speed graph changes to graphs involving the virt resource, it can
+eventually cause a panic. This is being tracked here:
+https://github.com/libvirt/libvirt-go/issues/7 (help welcome!)
+
+GLUSTERFS
+
+I've started working on some stuff for Glusterd so that it can embed
+mgmt as a lib, and use the resource model to simplify cluster
+management. This will involve writing Gluster resources in mgmt
+(volume, brick, etc...) and all sorts of other fun stuff. If you'd like
+to participate, please LMK!
+
+MISC
+
+We're still looking for new contributors, and there are easy, medium and hard issues available! You're also welcome to suggest your own! Please join us in #mgmtconfig on Freenode IRC, or ping this list if you'd like help getting started!
+
+THANKS
+
+We had 12 unique committers since 0.0.8, and have had 29 overall.
+Thanks (alphabetically) to everyone who contributed to the latest
+release: Daniele Sluijters, Daniel P. Berrange, Felix Frank, goberghen,
+James Shubin, Julien Pivotto, Kaushal M, Lars Kulseng, Mildred Ki'Lya,
+Sean Jones, Steve Milner, Tom Ritserveldt.
+
+Happy hacking,
+James
+@purpleidea

docs/resource-guide.md

@@ -60,10 +60,7 @@ it is not specified, but others cannot, and some might poorly infer if the
 struct name is ambiguous.
 
 If you'd like your resource to be accessible by the `YAML` graph API (GAPI),
-then you'll need to include the appropriate YAML fields as shown below. This is
-used by the `Puppet` compiler as well, so make sure you include these struct
-tags if you want existing `Puppet` code to be able to run using the `mgmt`
-engine.
+then you'll need to include the appropriate YAML fields as shown below.
 
 #### Example
 
@@ -96,14 +93,16 @@ Default() engine.Res
 ```
 
 This returns a populated resource struct as a `Res`. It shouldn't populate any
-values which already have the correct default as the golang zero value. In
+values which already get a good default as the respective golang zero value. In
 general it is preferable if the zero values make for the correct defaults.
+(This is to say, resources are designed to behave safely and intuitively
+when parameters take a zero value, whenever this is possible.)
 
 #### Example
 
 ```golang
 // Default returns some sensible defaults for this resource.
-func (obj *FooRes) Default() Res {
+func (obj *FooRes) Default() engine.Res {
 	return &FooRes{
 		Answer: 42, // sometimes, defaults shouldn't be the zero value
 	}
@@ -175,10 +174,10 @@ this. In other words, you should expect `Validate` to have run first, but you
 shouldn't allow `Init` to dangerously `rm -rf /$the_world` if your code only
 checks `$the_world` in `Validate`. Remember to always program safely!
 
-### Close
+### Cleanup
 
 ```golang
-Close() error
+Cleanup() error
 ```
 
 This is called to cleanup after the resource. It is usually not necessary, but
@@ -190,8 +189,8 @@ loop.
 #### Example
 
 ```golang
-// Close runs some cleanup code for this resource.
-func (obj *FooRes) Close() error {
+// Cleanup is run by the engine to clean up after the resource is done.
+func (obj *FooRes) Cleanup() error {
 	err := obj.conn.Close() // close some internal connection
 	obj.someMap = nil       // free up some large data structure from memory
 	return err
@@ -204,7 +203,7 @@ on an error if something went wrong.
 ### CheckApply
 
 ```golang
-CheckApply(apply bool) (checkOK bool, err error)
+CheckApply(ctx context.Context, apply bool) (checkOK bool, err error)
 ```
 
 `CheckApply` is where the real _work_ is done. Under normal circumstances, this
@@ -213,7 +212,8 @@ should return: `(true, nil)`. If the `apply` variable is set to `true`, then
 this means that we should then proceed to run the changes required to bring the
 resource into the correct state. If the `apply` variable is set to `false`, then
 the resource is operating in _noop_ mode and _no operational changes_ should be
-made!
+made! The ctx should be monitored in case a shutdown has been requested. This
+may be used if a timeout occurred, or if the user shutdown the engine.
 
 After having executed the necessary operations to bring the resource back into
 the desired state, or after having detected that the state was incorrect, but
@@ -232,7 +232,7 @@ to `CheckApply`.
 
 ```golang
 // CheckApply does the idempotent work of checking and applying resource state.
-func (obj *FooRes) CheckApply(apply bool) (bool, error) {
+func (obj *FooRes) CheckApply(ctx context.Context, apply bool) (bool, error) {
 	// check the state
 	if state_is_okay { return true, nil } // done early! :)
 
@@ -275,7 +275,7 @@ will likely find the state to now be correct.
 ### Watch
 
 ```golang
-Watch() error
+Watch(ctx context.Context) error
 ```
 
 `Watch` is a main loop that runs and sends messages when it detects that the
@@ -302,26 +302,25 @@ If the resource is activated in `polling` mode, the `Watch` method will not get
 executed. As a result, the resource must still work even if the main loop is not
 running.
 
+You must make sure to cleanup any running code or goroutines before Watch exits.
+
 #### Select
 
 The lifetime of most resources `Watch` method should be spent in an infinite
 loop that is bounded by a `select` call. The `select` call is the point where
 our method hands back control to the engine (and the kernel) so that we can
-sleep until something of interest wakes us up. In this loop we must process
-events from the engine via the `<-obj.init.Events` channel, and receive events
-for our resource itself!
+sleep until something of interest wakes us up. In this loop we must wait until
+we get a shutdown event from the engine via the `<-ctx.Done()` channel, which
+closes when we'd like to shut everything down. At this point you should cleanup,
+and let `Watch` close.
 
 #### Events
 
-If we receive an internal event from the `<-obj.init.Events` channel, we should
-read it with the `obj.init.Read` helper function. This function tells us if we
-should shutdown our resource. It also handles pause functionality which blocks
-our resource temporarily in this method. If this channel shuts down, then we
-should treat that as an exit signal.
-
-When we want to send an event, we use the `Event` helper function. It is also
-important to mark the resource state as `dirty` if we believe it might have
-changed. We do this by calling the `obj.init.Dirty` function.
+If the `<-ctx.Done()` channel closes, we should shutdown our resource. When we
+want to send an event, we use the `Event` helper function. This automatically
+marks the resource state as `dirty`. If you're unsure, it's not harmful to send
+the event. This will ultimately cause `CheckApply` to run. This method can block
+if the resource is being paused.
 
 #### Startup
 
@@ -330,8 +329,7 @@ to generate one event to notify the `mgmt` engine that we're now listening
 successfully, so that it can run an initial `CheckApply` to ensure we're safely
 tracking a healthy state and that we didn't miss anything when `Watch` was down
 or from before `mgmt` was running. You must do this by calling the
-`obj.init.Running` method. If it returns an error, you must exit and return that
-error.
+`obj.init.Running` method.
 
 #### Converged
 
@@ -349,7 +347,7 @@ sending out erroneous `Event` messages to keep things alive until it finishes.
 
 ```golang
 // Watch is the listener and main loop for this resource.
-func (obj *FooRes) Watch() error {
+func (obj *FooRes) Watch(ctx context.Context) error {
 	// setup the Foo resource
 	var err error
 	if err, obj.foo = OpenFoo(); err != nil {
@@ -358,41 +356,29 @@ func (obj *FooRes) Watch() error {
 	defer obj.whatever.CloseFoo() // shutdown our Foo
 
 	// notify engine that we're running
-	if err := obj.init.Running(); err != nil {
-		return err // exit if requested
-	}
+	obj.init.Running() // when started, notify engine that we're running
 
 	var send = false // send event?
 	for {
 		select {
-		case event, ok := <-obj.init.Events:
-			if !ok {
-				// shutdown engine
-				// (it is okay if some `defer` code runs first)
-				return nil
-			}
-			if err := obj.init.Read(event); err != nil {
-				return err
-			}
-
 		// the actual events!
 		case event := <-obj.foo.Events:
 			if is_an_event {
 				send = true
-				obj.init.Dirty() // dirty
 			}
 
 		// event errors
 		case err := <-obj.foo.Errors:
 			return err // will cause a retry or permanent failure
+
+		case <-ctx.Done(): // signal for shutdown request
+			return nil
 		}
 
 		// do all our event sending all together to avoid duplicate msgs
 		if send {
 			send = false
-			if err := obj.init.Event(); err != nil {
-				return err // exit if requested
-			}
+			obj.init.Event()
 		}
 	}
 }
@@ -567,24 +553,6 @@ ready to detect changes.
 Event sends an event notifying the engine of a possible state change. It is
 only called from within `Watch`.
 
-### Events
-
-Events is a channel that we must watch for messages from the engine. When it
-closes, this is a signal to shutdown. It is
-only called from within `Watch`.
-
-### Read
-
-Read processes messages that come in from the `Events` channel. It is a helper
-method that knows how to handle the pause mechanism correctly. It is
-only called from within `Watch`.
-
-### Dirty
-
-Dirty marks the resource state as dirty. This signals to the engine that
-CheckApply will have some work to do in order to converge it. It is
-only called from within `Watch`.
-
 ### Refresh
 
 Refresh returns whether the resource received a notification. This flag can be
@@ -609,15 +577,15 @@ It is only called from within `CheckApply`.
 World provides a connection to the outside world. This is most often used for
 communicating with the distributed database. It can be used in `Init`,
 `CheckApply` and `Watch`. Use with discretion and understanding of the internals
-if needed in `Close`.
+if needed in `Cleanup`.
 
 ### VarDir
 
 VarDir is a facility for local storage. It is used to return a path to a
 directory which may be used for temporary storage. It should be cleaned up on
-resource `Close` if the resource would like to delete the contents. The resource
-should not assume that the initial directory is empty, and it should be cleaned
-on `Init` if that is a requirement.
+resource `Cleanup` if the resource would like to delete the contents. The
+resource should not assume that the initial directory is empty, and it should be
+cleaned on `Init` if that is a requirement.
 
 ### Debug
 
@@ -652,7 +620,7 @@ func init() { // special golang method that runs once
 
 To support YAML unmarshalling for your resource, you must implement an
 additional method. It is recommended if you want to use your resource with the
-`Puppet` compiler.
+`yaml` compiler.
 
 ```golang
 UnmarshalYAML(unmarshal func(interface{}) error) error // optional
@@ -669,8 +637,8 @@ The signature intentionally matches what is required to satisfy the `go-yaml`
 #### Example
 
 ```golang
-// UnmarshalYAML is the custom unmarshal handler for this struct.
-// It is primarily useful for setting the defaults.
+// UnmarshalYAML is the custom unmarshal handler for this struct. It is
+// primarily useful for setting the defaults.
 func (obj *FooRes) UnmarshalYAML(unmarshal func(interface{}) error) error {
 	type rawRes FooRes // indirection to avoid infinite recursion
 
@@ -707,10 +675,10 @@ receiving one. This can _only_ be done inside of the `CheckApply` function!
 
 ```golang
 // inside CheckApply, probably near the top
-if val, exists := obj.init.Recv()["SomeKey"]; exists {
-	obj.init.Logf("the SomeKey param was sent to us from: %s.%s", val.Res, val.Key)
+if val, exists := obj.init.Recv()["some_key"]; exists {
+	obj.init.Logf("the some_key param was sent to us from: %s.%s", val.Res, val.Key)
 	if val.Changed {
-		obj.init.Logf("the SomeKey param was just updated!")
+		obj.init.Logf("the some_key param was just updated!")
 		// you may want to invalidate some local cache
 	}
 }
@@ -754,16 +722,16 @@ two calls, this is much more difficult. A common example is that a resource
 might want to open a connection to `dbus` or `http` to do resource state testing
 and applying. If the methods are combined, there's no need to open and close
 them twice. A counter argument might be that you could open the connection in
-`Init`, and close it in `Close`, however you might not want that open for the
+`Init`, and close it in `Cleanup`, however you might not want that open for the
 full lifetime of the resource if you only change state occasionally.
 2. Suppose you came up with a really good reason why you wanted the two methods
 to be separate. It turns out that the current `CheckApply` can wrap this easily.
 It would look approximately like this:
 
 ```golang
-func (obj *FooRes) CheckApply(apply bool) (bool, error) {
+func (obj *FooRes) CheckApply(ctx context.Context, apply bool) (bool, error) {
 	// my private split implementation of check and apply
-	if c, err := obj.check(); err != nil {
+	if c, err := obj.check(ctx); err != nil {
 		return false, err // we errored
 	} else if c {
 		return true, nil // state was good!
@@ -773,7 +741,7 @@ func (obj *FooRes) CheckApply(apply bool) (bool, error) {
 		return false, nil // state needs fixing, but apply is false
 	}
 
-	err := obj.apply() // errors if failure or unable to apply
+	err := obj.apply(ctx) // errors if failure or unable to apply
 
 	return false, err // always return false, with an optional error
 }
@@ -783,6 +751,23 @@ Feel free to use this pattern if you're convinced it's necessary. Alternatively,
 if you think I got the `Res` API wrong and you have an improvement, please let
 us know!
 
+### Why do resources have both a `Cmp` method and an `IFF` (on the UID) method?
+
+The `Cmp()` methods are for determining if two resources are effectively the
+same, which is used to make graph change delta's efficient. This is when we want
+to change from the current running graph to a new graph, but preserve the common
+vertices. Since we want to make this process efficient, we only update the parts
+that are different, and leave everything else alone. This `Cmp()` method can
+tell us if two resources are the same. In case it is not obvious, `cmp` is an
+abbrev. for compare.
+
+The `IFF()` method is part of the whole UID system, which is for discerning if a
+resource meets the requirements another expects for an automatic edge. This is
+because the automatic edge system assumes a unified UID pattern to test for
+equality. In the future it might be helpful or sane to merge the two similar
+comparison functions although for now they are separate because they are
+actually answer different questions.
+
 ### What new resource primitives need writing?
 
 There are still many ideas for new resources that haven't been written yet. If

docs/resources.md

@@ -17,6 +17,7 @@ You might want to look at the [generated documentation](https://godoc.org/github
 for more up-to-date information about these resources.
 
 * [Augeas](#Augeas): Manipulate files using augeas.
+* [Consul:KV](#ConsulKV): Set keys in a Consul datastore.
 * [Docker](#Docker):[Container](#Container) Manage docker containers.
 * [Exec](#Exec): Execute shell commands on the system.
 * [File](#File): Manage files and directories.
@@ -32,6 +33,8 @@ for more up-to-date information about these resources.
 * [Print](#Print): Print messages to the console.
 * [Svc](#Svc): Manage system systemd services.
 * [Test](#Test): A mostly harmless resource that is used for internal testing.
+* [Tftp:File](#TftpFile): Add files to the small embedded embedded tftp server.
+* [Tftp:Server](#TftpServer): Run a small embedded tftp server.
 * [Timer](#Timer): Manage system systemd services.
 * [User](#User): Manage system users.
 * [Virt](#Virt): Manage virtual machines with libvirt.
@@ -68,10 +71,10 @@ identified by a trailing slash in their path name. File have no such slash.
 
 It has the following properties:
 
-* `path`: file path (directories have a trailing slash here)
+* `path`: absolute file path (directories have a trailing slash here)
+* `state`: either `exists`, `absent`, or undefined
 * `content`: raw file content
-* `state`: either `exists` (the default value) or `absent`
-* `mode`: octal unix file permissions
+* `mode`: octal unix file permissions or symbolic string
 * `owner`: username or uid for the file owner
 * `group`: group name or gid for the file group
 
@@ -79,6 +82,16 @@ It has the following properties:
 
 The path property specifies the file or directory that we are managing.
 
+### State
+
+The state property describes the action we'd like to apply for the resource. The
+possible values are: `exists` and `absent`. If you do not specify either of
+these, it is undefined. Without specifying this value as `exists`, another param
+cannot cause a file to get implicitly created. When specifying this value as
+`absent`, you should not specify any other params that would normally change the
+file. For example, if you specify `content` and this param is `absent`, then you
+will get an engine validation error.
+
 ### Content
 
 The content property is a string that specifies the desired file contents.
@@ -88,10 +101,12 @@ The content property is a string that specifies the desired file contents.
 The source property points to a source file or directory path that we wish to
 copy over and use as the desired contents for our resource.
 
-### State
+### Fragments
 
-The state property describes the action we'd like to apply for the resource. The
-possible values are: `exists` and `absent`.
+The fragments property lets you specify a list of files to concatenate together
+to make up the contents of this file. They will be combined in the order that
+they are listed in. If one of the files specified is a directory, then the
+files in that top-level directory will be themselves combined together and used.
 
 ### Recurse
 
@@ -104,6 +119,12 @@ The force property is required if we want the file resource to be able to change
 a file into a directory or vice-versa. If such a change is needed, but the force
 property is not set to `true`, then this file resource will error.
 
+### Purge
+
+The purge property is used when this file represents a directory, and we'd like
+to remove any unmanaged files from within it. Please note that any unmanaged
+files in a directory with this flag set will be irreversibly deleted.
+
 ## Group
 
 The group resource manages the system groups from `/etc/group`.
@@ -206,6 +227,16 @@ The service resource is still very WIP. Please help us by improving it!
 
 The test resource is mostly harmless and is used for internal tests.
 
+## Tftp:File
+
+This adds files to the running tftp server. It's useful because it allows you to
+add individual files without needing to create them on disk.
+
+## Tftp:Server
+
+Run a small embedded tftp server. This doesn't apply any state, but instead runs
+a pure golang tftp server in the Watch loop.
+
 ## Timer
 
 This resource needs better documentation. Please help us by improving it!

docs/style-guide.md

@@ -1,22 +1,28 @@
 # Style guide
 
-## Overview
+This document aims to be a reference for the desired style for patches to mgmt,
+and the associated `mcl` language. In particular it describes conventions which
+are not officially enforced by tools and in test cases, or that aren't clearly
+defined elsewhere. We try to turn as many of these into automated tests as we
+can. If something here is not defined in a test, or you think it should be,
+please write one! Even better, you can write a tool to automatically fix it,
+since this is more useful and can easily be turned into a test!
 
-This document aims to be a reference for the desired style for patches to mgmt.
-In particular it describes conventions which we use which are not officially
-enforced by the `gofmt` tool, and which might not be clearly defined elsewhere.
-Most of these are common sense to seasoned programmers, and we hope this will be
-a useful reference for new programmers.
+## Overview for golang code
+
+Most style issues are enforced by the `gofmt` tool. Other style aspects are
+often common sense to seasoned programmers, and we hope this will be a useful
+reference for new programmers.
 
 There are a lot of useful code review comments described
 [here](https://github.com/golang/go/wiki/CodeReviewComments). We don't
 necessarily follow everything strictly, but it is in general a very good guide.
 
-## Basics
+### Basics
 
 * All of our golang code is formatted with `gofmt`.
 
-## Comments
+### Comments
 
 All of our code is commented with the minimums required for `godoc` to function,
 and so that our comments pass `golint`. Code comments should either be full
@@ -28,7 +34,7 @@ They should explain algorithms, describe non-obvious behaviour, or situations
 which would otherwise need explanation or additional research during a code
 review. Notes about use of unfamiliar API's is a good idea for a code comment.
 
-### Example
+#### Example
 
 Here you can see a function with the correct `godoc` string. The first word must
 match the name of the function. It is _not_ capitalized because the function is
@@ -41,7 +47,7 @@ func square(x int) int {
 }
 ```
 
-## Line length
+### Line length
 
 In general we try to stick to 80 character lines when it is appropriate. It is
 almost *always* appropriate for function `godoc` comments and most longer
@@ -55,7 +61,13 @@ Occasionally inline, two line source code comments are used within a function.
 These should usually be balanced so that you don't have one line with 78
 characters and the second with only four. Split the comment between the two.
 
-## Method receiver naming
+### Default values
+
+Whenever a constant or function parameter is defined, try and have the safer or
+default value be the `zero` value. For example, instead of `const NoDanger`, use
+`const AllowDanger` so that the `false` value is the safe scenario.
+
+### Method receiver naming
 
 [Contrary](https://github.com/golang/go/wiki/CodeReviewComments#receiver-names)
 to the specialized naming of the method receiver variable, we usually name all
@@ -65,7 +77,7 @@ makes the code easier to read since you don't need to remember the name of the
 method receiver variable in each different method. This is very similar to what
 is done in `python`.
 
-### Example
+#### Example
 
 ```golang
 // Bar does a thing, and returns the number of baz results found in our
@@ -78,7 +90,76 @@ func (obj *Foo) Bar(baz string) int {
 }
 ```
 
-## Consistent ordering
+### Variable naming
+
+We prefer shorter, scoped variables rather than `unnecessarilyLongIdentifiers`.
+Remember the scoping rules and feel free to use new variables where appropriate.
+For example, in a short string snippet you can use `s` instead of `myString`, as
+well as other common choices. `i` is a common `int` counter, `f` for files, `fn`
+for functions, `x` for something else and so on.
+
+### Variable re-use
+
+Feel free to create and use new variables instead of attempting to re-use the
+same string. For example, if a function input arg is named `s`, you can use a
+new variable to receive the first computation result on `s` instead of storing
+it back into the original `s`. This avoids confusion if a different part of the
+code wants to read the original input, and it avoids any chance of edit by
+reference of the original callers copy of the variable.
+
+#### Example
+
+```golang
+MyNotIdealFunc(s string, b bool) string {
+	if !b {
+		return s + "hey"
+	}
+	s = strings.Replace(s, "blah", "", -1) // not ideal (re-use of `s` var)
+	return s
+}
+
+MyOkayFunc(s string, b bool) string {
+	if !b {
+		return s + "hey"
+	}
+	s2 := strings.Replace(s, "blah", "", -1) // doesn't re-use `s` variable
+	return s2
+}
+
+MyGreatFunc(s string, b bool) string {
+	if !b {
+		return s + "hey"
+	}
+	return strings.Replace(s, "blah", "", -1) // even cleaner
+}
+```
+
+### Constants in code
+
+If a function takes a specifier (often a bool) it's sometimes better to name
+that variable (often with a `const`) rather than leaving a naked `bool` in the
+code. For example, `x := MyFoo("blah", false)` is less clear than
+`const useMagic = false; x := MyFoo("blah", useMagic)`.
+
+### Empty slice declarations
+
+When declaring a new empty slice, there are three different mechanisms:
+
+1. `a := []string{}`
+
+2. `var a []string`
+
+3. `a := make([]string, 0)`
+
+In general, we prefer the first method because we find that it is succinct, and
+very readable. The third method is the least recommended because you're adding
+extra data that a smart compiler could probably figure out on its own. There are
+performance implications between these three methods, so unless your code is in
+a fast path or memory constrained environment where this matters (and that you
+ideally have proof of this) please use the methods as ordered as much as
+possible.
+
+### Consistent ordering
 
 In general we try to preserve a logical ordering in source files which usually
 matches the common order of execution that a _lazy evaluator_ would follow.
@@ -90,6 +171,72 @@ declared in the interface.
 When implementing code for the various types in the language, please follow this
 order: `bool`, `str`, `int`, `float`, `list`, `map`, `struct`, `func`.
 
+For other aspects where you have a set of items, try to be internally consistent
+as well. For example, if you have two switch statements with `A`, `B`, and `C`,
+please use the same ordering for these elements elsewhere that they appear in
+the code and in the commentary if it is not illogical to do so.
+
+### Product identifiers
+
+Try to avoid references in the code to `mgmt` or a specific program name string
+if possible. This makes it easier to rename code if we ever pick a better name
+or support `libmgmt` better if we embed it. You can use the `Program` variable
+which is available in numerous places if you want a string to put in the logs.
+
+It is also recommended to avoid the `go` (programming language name) string if
+possible. Try to use `golang` if required, since the word `go` is already
+overloaded, and in particular it was even already used by the
+[`go!`](https://en.wikipedia.org/wiki/Go!_(programming_language)).
+
+## Overview for mcl code
+
+The `mcl` language is quite new, so this guide will probably change over time as
+we find what's best, and hopefully we'll be able to add an `mclfmt` tool in the
+future so that less of this needs to be documented. (Patches welcome!)
+
+### Indentation
+
+Code indentation is done with tabs. The tab-width is a private preference, which
+is the beauty of using tabs: you can have your own personal preference. The
+inventor of `mgmt` uses and recommends a width of eight, and that is what should
+be used if your tool requires a modeline to be publicly committed.
+
+### Line length
+
+We recommend you stick to 80 char line width. If you find yourself with deeper
+nesting, it might be a hint that your code could be refactored in a more
+pleasant way.
+
+### Capitalization
+
+At the moment, variables, function names, and classes are all lowercase and do
+not contain underscores. We will probably figure out what style to recommend
+when the language is a bit further along. For example, we haven't decided if we
+should have a notion of public and private variables, and if we'd like to
+reserve capitalization for this situation.
+
+### Module naming
+
+We recommend you name your modules with an `mgmt-` prefix. For example, a module
+about bananas might be named `mgmt-banana`. This is helpful for the useful magic
+built-in to the module import code, which will by default take a remote import
+like: `import "https://github.com/purpleidea/mgmt-banana/"` and namespace it as
+`banana`. Of course you can always pick the namespace yourself on import with:
+`import "https://github.com/purpleidea/mgmt-banana/" as tomato` or something
+similar.
+
+### Licensing
+
+We believe that sharing code helps reduce unnecessary re-invention, so that we
+can [stand on the shoulders of giants](https://en.wikipedia.org/wiki/Standing_on_the_shoulders_of_giants)
+and hopefully make faster progress in science, medicine, exploration, etc... As
+a result, we recommend releasing your modules under the [LGPLv3+](https://www.gnu.org/licenses/lgpl-3.0.en.html)
+license for the maximum balance of freedom and re-usability. We strongly oppose
+any [CLA](https://en.wikipedia.org/wiki/Contributor_License_Agreement)
+requirements and believe that the ["inbound==outbound"](https://ref.fedorapeople.org/fontana-linuxcon.html#slide2)
+rule applies. Lastly, we do not support software patents and we hope you don't
+either!
+
 ## Suggestions
 
 If you have any ideas for suggestions or other improvements to this guide,

engine/autoedge.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,7 +13,19 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package engine
 
@@ -31,6 +43,10 @@ type EdgeableRes interface {
 	// trait.
 	AutoEdgeMeta() *AutoEdgeMeta
 
+	// SetAutoEdgeMeta lets you set all of the meta params for the automatic
+	// edges trait in a single call.
+	SetAutoEdgeMeta(*AutoEdgeMeta)
+
 	// UIDs includes all params to make a unique identification of this
 	// object.
 	UIDs() []ResUID // most resources only return one
@@ -64,7 +80,8 @@ type AutoEdge interface {
 	Test([]bool) bool // call until false
 }
 
-// ResUID is a unique identifier for a resource, namely it's name, and the kind ("type").
+// ResUID is a unique identifier for a resource, namely it's name, and the kind
+// ("type").
 type ResUID interface {
 	fmt.Stringer // String() string
 
@@ -100,9 +117,9 @@ func (obj *BaseUID) String() string {
 }
 
 // IFF looks at two UID's and if and only if they are equivalent, returns true.
-// If they are not equivalent, it returns false.
-// Most resources will want to override this method, since it does the important
-// work of actually discerning if two resources are identical in function.
+// If they are not equivalent, it returns false. Most resources will want to
+// override this method, since it does the important work of actually discerning
+// if two resources are identical in function.
 func (obj *BaseUID) IFF(uid ResUID) bool {
 	res, ok := uid.(*BaseUID)
 	if !ok {

engine/autoedge_test.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,9 +13,21 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
-// +build !root
+//go:build !root
 
 package engine
 

engine/autogroup.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,7 +13,19 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package engine
 
@@ -34,11 +46,16 @@ type GroupableRes interface {
 	// grouping trait.
 	AutoGroupMeta() *AutoGroupMeta
 
+	// SetAutoGroupMeta lets you set all of the meta params for the
+	// automatic grouping trait in a single call.
+	SetAutoGroupMeta(*AutoGroupMeta)
+
 	// GroupCmp compares two resources and decides if they're suitable for
-	//grouping. This usually needs to be unique to your resource.
+	// grouping. This usually needs to be unique to your resource.
 	GroupCmp(res GroupableRes) error
 
-	// GroupRes groups resource argument (res) into self.
+	// GroupRes groups resource argument (res) into self. Callers of this
+	// method should probably also run SetParent.
 	GroupRes(res GroupableRes) error
 
 	// IsGrouped determines if we are grouped.
@@ -50,8 +67,15 @@ type GroupableRes interface {
 	// GetGroup returns everyone grouped inside me.
 	GetGroup() []GroupableRes // return everyone grouped inside me
 
-	// SetGroup sets the grouped resources into me.
+	// SetGroup sets the grouped resources into me. Callers of this method
+	// should probably also run SetParent.
 	SetGroup([]GroupableRes)
+
+	// Parent returns the parent groupable resource that I am inside of.
+	Parent() GroupableRes
+
+	// SetParent tells a particular grouped resource who their parent is.
+	SetParent(res GroupableRes)
 }
 
 // AutoGroupMeta provides some parameters specific to automatic grouping.

engine/cmp.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,7 +13,19 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package engine
 
@@ -21,10 +33,12 @@ import (
 	"fmt"
 
 	"github.com/purpleidea/mgmt/pgraph"
+	"github.com/purpleidea/mgmt/util/errwrap"
 )
 
 // ResCmp compares two resources by checking multiple aspects. This is the main
-// entry point for running all the compare steps on two resource.
+// entry point for running all the compare steps on two resources. This code is
+// very similar to AdaptCmp.
 func ResCmp(r1, r2 Res) error {
 	if r1.Kind() != r2.Kind() {
 		return fmt.Errorf("kind differs")
@@ -37,6 +51,30 @@ func ResCmp(r1, r2 Res) error {
 		return err
 	}
 
+	// TODO: do we need to compare other traits/metaparams?
+
+	m1 := r1.MetaParams()
+	m2 := r2.MetaParams()
+	if (m1 == nil) != (m2 == nil) { // xor
+		return fmt.Errorf("meta params differ")
+	}
+	if m1 != nil && m2 != nil {
+		if err := m1.Cmp(m2); err != nil {
+			return err
+		}
+	}
+
+	r1x, ok1 := r1.(RefreshableRes)
+	r2x, ok2 := r2.(RefreshableRes)
+	if ok1 != ok2 {
+		return fmt.Errorf("refreshable differs") // they must be different (optional)
+	}
+	if ok1 && ok2 {
+		if r1x.Refresh() != r2x.Refresh() {
+			return fmt.Errorf("refresh differs")
+		}
+	}
+
 	// compare meta params for resources with auto edges
 	r1e, ok1 := r1.(EdgeableRes)
 	r2e, ok2 := r2.(EdgeableRes)
@@ -81,12 +119,243 @@ func ResCmp(r1, r2 Res) error {
 			for k := range ix {
 				// compare sub resources
 				if err := ResCmp(ix[k], jx[k]); err != nil {
+					//fmt.Printf("bad Cmp: %+v <> %+v for: %+v <> %+v err: %+v\n", r1, r2, ix[k], jx[k], err)
 					return err
 				}
 			}
 		}
 	}
 
+	r1r, ok1 := r1.(RecvableRes)
+	r2r, ok2 := r2.(RecvableRes)
+	if ok1 != ok2 {
+		return fmt.Errorf("recvable differs") // they must be different (optional)
+	}
+	if ok1 && ok2 {
+		v1 := r1r.Recv()
+		v2 := r2r.Recv()
+
+		// XXX: Our Send/Recv in the lib/main.go doesn't seem to be
+		// pulling this in, so this always compares differently. We can
+		// comment it out for now, since it's not too consequential.
+		// XXX: Find out what the issue is and fix it for here and send.
+		// XXX: The below errors are commented out until this is fixed.
+		if (v1 == nil) != (v2 == nil) { // xor
+			//return fmt.Errorf("recv params differ")
+		}
+		if v1 != nil && v2 != nil {
+			if len(v1) != len(v2) {
+				//return fmt.Errorf("recv param lengths differ")
+			}
+			for key, send1 := range v1 { // map[string]*engine.Send
+				send2, exists := v2[key]
+				if !exists {
+					//return fmt.Errorf("recv param key %s doesn't exist", key)
+				}
+				if (send1 == nil) != (send2 == nil) { // xor
+					//return fmt.Errorf("recv param key %s send differs", key)
+				}
+				if send1 != nil && send2 != nil && send1.Key != send2.Key {
+					//return fmt.Errorf("recv param key %s send key differs (%v != %v)", key, send1.Key, send2.Key)
+				}
+			}
+		}
+	}
+
+	r1s, ok1 := r1.(SendableRes)
+	r2s, ok2 := r2.(SendableRes)
+	if ok1 != ok2 {
+		return fmt.Errorf("sendable differs") // they must be different (optional)
+	}
+	if ok1 && ok2 {
+		s1 := r1s.Sent()
+		s2 := r2s.Sent()
+
+		// XXX: Our Send/Recv in the lib/main.go doesn't seem to be
+		// pulling this in, so this always compares differently. We can
+		// comment it out for now, since it's not too consequential.
+		// XXX: Find out what the issue is and fix it for here and recv.
+		// XXX: The below errors are commented out until this is fixed.
+		if (s1 == nil) != (s2 == nil) { // xor
+			//return fmt.Errorf("send params differ")
+		}
+		if s1 != nil && s2 != nil {
+			// TODO: reflect.DeepEqual?
+			//return fmt.Errorf("send params exist")
+		}
+	}
+
+	// compare meta params for resources with reversible traits
+	r1v, ok1 := r1.(ReversibleRes)
+	r2v, ok2 := r2.(ReversibleRes)
+	if ok1 != ok2 {
+		return fmt.Errorf("reversible differs") // they must be different (optional)
+	}
+	if ok1 && ok2 {
+		if r1v.ReversibleMeta().Cmp(r2v.ReversibleMeta()) != nil {
+			return fmt.Errorf("reversible differs")
+		}
+	}
+
+	return nil
+}
+
+// AdaptCmp compares two resources by checking multiple aspects. This is the
+// main entry point for running all the compatible compare steps on two
+// resources. This code is very similar to ResCmp.
+func AdaptCmp(r1, r2 CompatibleRes) error {
+	if r1.Kind() != r2.Kind() {
+		return fmt.Errorf("kind differs")
+	}
+	if r1.Name() != r2.Name() {
+		return fmt.Errorf("name differs")
+	}
+
+	// run `Adapts` instead of `Cmp`
+	if err := r1.Adapts(r2); err != nil {
+		return err
+	}
+
+	// TODO: do we need to compare other traits/metaparams?
+
+	m1 := r1.MetaParams()
+	m2 := r2.MetaParams()
+	if (m1 == nil) != (m2 == nil) { // xor
+		return fmt.Errorf("meta params differ")
+	}
+	if m1 != nil && m2 != nil {
+		if err := m1.Cmp(m2); err != nil {
+			return err
+		}
+	}
+
+	// we don't need to compare refresh, since those can always be merged...
+
+	// compare meta params for resources with auto edges
+	r1e, ok1 := r1.(EdgeableRes)
+	r2e, ok2 := r2.(EdgeableRes)
+	if ok1 != ok2 {
+		return fmt.Errorf("edgeable differs") // they must be different (optional)
+	}
+	if ok1 && ok2 {
+		if r1e.AutoEdgeMeta().Cmp(r2e.AutoEdgeMeta()) != nil {
+			return fmt.Errorf("autoedge differs")
+		}
+	}
+
+	// compare meta params for resources with auto grouping
+	r1g, ok1 := r1.(GroupableRes)
+	r2g, ok2 := r2.(GroupableRes)
+	if ok1 != ok2 {
+		return fmt.Errorf("groupable differs") // they must be different (optional)
+	}
+	if ok1 && ok2 {
+		if r1g.AutoGroupMeta().Cmp(r2g.AutoGroupMeta()) != nil {
+			return fmt.Errorf("autogroup differs")
+		}
+
+		// if resources are grouped, are the groups the same?
+		if i, j := r1g.GetGroup(), r2g.GetGroup(); len(i) != len(j) {
+			return fmt.Errorf("autogroup groups differ")
+		} else if len(i) > 0 { // trick the golinter
+
+			// Sort works with Res, so convert the lists to that
+			iRes := []Res{}
+			for _, r := range i {
+				res := r.(Res)
+				iRes = append(iRes, res)
+			}
+			jRes := []Res{}
+			for _, r := range j {
+				res := r.(Res)
+				jRes = append(jRes, res)
+			}
+
+			ix, jx := Sort(iRes), Sort(jRes) // now sort :)
+			for k := range ix {
+				// compare sub resources
+				// TODO: should we use AdaptCmp here?
+				// TODO: how would they run `Merge` ? (we don't)
+				// this code path will probably not run, because
+				// it is called in the lang before autogrouping!
+				if err := ResCmp(ix[k], jx[k]); err != nil {
+					return err
+				}
+			}
+		}
+	}
+
+	r1r, ok1 := r1.(RecvableRes)
+	r2r, ok2 := r2.(RecvableRes)
+	if ok1 != ok2 {
+		return fmt.Errorf("recvable differs") // they must be different (optional)
+	}
+	if ok1 && ok2 {
+		v1 := r1r.Recv()
+		v2 := r2r.Recv()
+
+		// XXX: Our Send/Recv in the lib/main.go doesn't seem to be
+		// pulling this in, so this always compares differently. We can
+		// comment it out for now, since it's not too consequential.
+		// XXX: Find out what the issue is and fix it for here and send.
+		// XXX: The below errors are commented out until this is fixed.
+		if (v1 == nil) != (v2 == nil) { // xor
+			//return fmt.Errorf("recv params differ")
+		}
+		if v1 != nil && v2 != nil {
+			if len(v1) != len(v2) {
+				//return fmt.Errorf("recv param lengths differ")
+			}
+			for key, send1 := range v1 { // map[string]*engine.Send
+				send2, exists := v2[key]
+				if !exists {
+					//return fmt.Errorf("recv param key %s doesn't exist", key)
+				}
+				if (send1 == nil) != (send2 == nil) { // xor
+					//return fmt.Errorf("recv param key %s send differs", key)
+				}
+				if send1 != nil && send2 != nil && send1.Key != send2.Key {
+					//return fmt.Errorf("recv param key %s send key differs (%v != %v)", key, send1.Key, send2.Key)
+				}
+			}
+		}
+	}
+
+	r1s, ok1 := r1.(SendableRes)
+	r2s, ok2 := r2.(SendableRes)
+	if ok1 != ok2 {
+		return fmt.Errorf("sendable differs") // they must be different (optional)
+	}
+	if ok1 && ok2 {
+		s1 := r1s.Sent()
+		s2 := r2s.Sent()
+
+		// XXX: Our Send/Recv in the lib/main.go doesn't seem to be
+		// pulling this in, so this always compares differently. We can
+		// comment it out for now, since it's not too consequential.
+		// XXX: Find out what the issue is and fix it for here and recv.
+		// XXX: The below errors are commented out until this is fixed.
+		if (s1 == nil) != (s2 == nil) { // xor
+			//return fmt.Errorf("send params differ")
+		}
+		if s1 != nil && s2 != nil {
+			// TODO: reflect.DeepEqual?
+			//return fmt.Errorf("send params exist")
+		}
+	}
+
+	// compare meta params for resources with reversible traits
+	r1v, ok1 := r1.(ReversibleRes)
+	r2v, ok2 := r2.(ReversibleRes)
+	if ok1 != ok2 {
+		return fmt.Errorf("reversible differs") // they must be different (optional)
+	}
+	if ok1 && ok2 {
+		if r1v.ReversibleMeta().Cmp(r2v.ReversibleMeta()) != nil {
+			return fmt.Errorf("reversible differs")
+		}
+	}
+
 	return nil
 }
 
@@ -103,9 +372,11 @@ func VertexCmpFn(v1, v2 pgraph.Vertex) (bool, error) {
 		return false, fmt.Errorf("v2 is not a Res")
 	}
 
-	if ResCmp(r1, r2) != nil {
+	if err := ResCmp(r1, r2); err != nil {
+		//fmt.Printf("bad Cmp: %p %+v <> %p %+v err: %+v\n", r1, r1, r2, r2, err)
 		return false, nil
 	}
+	//fmt.Printf("ok Cmp: %p %+v <> %p %+v\n", r1, r1, r2, r2)
 
 	return true, nil
 }
@@ -124,3 +395,55 @@ func EdgeCmpFn(e1, e2 pgraph.Edge) (bool, error) {
 	}
 	return edge1.Cmp(edge2) == nil, nil
 }
+
+// ResGraphMapper compares two graphs, and gives us a mapping from new to old
+// based on the resource kind and name only. This allows us to know which
+// previous resource might have data to pass on to the new version in the next
+// generation.
+// FIXME: Optimize this for performance since it runs a lot...
+func ResGraphMapper(oldGraph, newGraph *pgraph.Graph) (map[RecvableRes]RecvableRes, error) {
+	mapper := make(map[RecvableRes]RecvableRes) // new -> old based on name and kind only?
+	cmp := func(r1, r2 Res) error {
+		if r1.Kind() != r2.Kind() {
+			return fmt.Errorf("kind differs")
+		}
+		if r1.Name() != r2.Name() {
+			return fmt.Errorf("name differs")
+		}
+		return nil
+	}
+
+	// XXX: run this as a topological sort or reverse topological sort?
+	for v := range newGraph.Adjacency() { // loop through the vertices (resources)
+		r, ok := v.(RecvableRes)
+		if !ok {
+			continue // skip
+		}
+		fn := func(vv pgraph.Vertex) (bool, error) {
+			rr, ok := vv.(Res)
+			if !ok {
+				return false, fmt.Errorf("not a Res")
+			}
+
+			if err := cmp(rr, r); err != nil {
+				return false, nil
+			}
+			return true, nil
+		}
+		vertex, err := oldGraph.VertexMatchFn(fn)
+		if err != nil {
+			return nil, errwrap.Wrapf(err, "VertexMatchFn failed")
+		}
+		if vertex == nil {
+			continue // skip (error?)
+		}
+		res, ok := vertex.(RecvableRes)
+		if !ok {
+			continue // skip (error?)
+		}
+
+		mapper[r] = res
+	}
+
+	return mapper, nil
+}

engine/copy.go

@@ -0,0 +1,183 @@
+// Mgmt
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
+
+package engine
+
+import (
+	"fmt"
+
+	"github.com/purpleidea/mgmt/util/errwrap"
+)
+
+// ResCopy copies a resource. This is the main entry point for copying a
+// resource since it does all the common engine-level copying as well.
+func ResCopy(r CopyableRes) (CopyableRes, error) {
+	res := r.Copy()
+	res.SetKind(r.Kind())
+	res.SetName(r.Name())
+
+	if x, ok := r.(MetaRes); ok {
+		dst, ok := res.(MetaRes)
+		if !ok {
+			// programming error
+			panic("meta interfaces are illogical")
+		}
+		dst.SetMetaParams(x.MetaParams().Copy()) // copy b/c we have it
+	}
+
+	if x, ok := r.(RefreshableRes); ok {
+		dst, ok := res.(RefreshableRes)
+		if !ok {
+			// programming error
+			panic("refresh interfaces are illogical")
+		}
+		dst.SetRefresh(x.Refresh()) // no need to copy atm
+	}
+
+	// copy meta params for resources with auto edges
+	if x, ok := r.(EdgeableRes); ok {
+		dst, ok := res.(EdgeableRes)
+		if !ok {
+			// programming error
+			panic("autoedge interfaces are illogical")
+		}
+		dst.SetAutoEdgeMeta(x.AutoEdgeMeta()) // no need to copy atm
+	}
+
+	// copy meta params for resources with auto grouping
+	if x, ok := r.(GroupableRes); ok {
+		dst, ok := res.(GroupableRes)
+		if !ok {
+			// programming error
+			panic("autogroup interfaces are illogical")
+		}
+		dst.SetAutoGroupMeta(x.AutoGroupMeta()) // no need to copy atm
+
+		grouped := []GroupableRes{}
+		for _, g := range x.GetGroup() {
+			g0, ok := g.(CopyableRes)
+			if !ok {
+				return nil, fmt.Errorf("resource wasn't copyable")
+			}
+			g1, err := ResCopy(g0)
+			if err != nil {
+				return nil, err
+			}
+			g2, ok := g1.(GroupableRes)
+			if !ok {
+				return nil, fmt.Errorf("resource wasn't groupable")
+			}
+			g2.SetParent(dst) // store who my parent is
+			grouped = append(grouped, g2)
+		}
+		dst.SetGroup(grouped)
+	}
+
+	if x, ok := r.(RecvableRes); ok {
+		dst, ok := res.(RecvableRes)
+		if !ok {
+			// programming error
+			panic("recv interfaces are illogical")
+		}
+		dst.SetRecv(x.Recv()) // no need to copy atm
+	}
+
+	if x, ok := r.(SendableRes); ok {
+		dst, ok := res.(SendableRes)
+		if !ok {
+			// programming error
+			panic("send interfaces are illogical")
+		}
+		if err := dst.Send(x.Sent()); err != nil { // no need to copy atm
+			return nil, errwrap.Wrapf(err, "can't copy send")
+		}
+	}
+
+	// copy meta params for resources with reversible traits
+	if x, ok := r.(ReversibleRes); ok {
+		dst, ok := res.(ReversibleRes)
+		if !ok {
+			// programming error
+			panic("reversible interfaces are illogical")
+		}
+		dst.SetReversibleMeta(x.ReversibleMeta()) // no need to copy atm
+	}
+
+	return res, nil
+}
+
+// ResMerge merges a set of resources that are compatible with each other. This
+// is the main entry point for the merging. They must each successfully be able
+// to run AdaptCmp without error.
+func ResMerge(r ...CompatibleRes) (CompatibleRes, error) {
+	if len(r) == 0 {
+		return nil, fmt.Errorf("zero resources given")
+	}
+	if len(r) == 1 {
+		return r[0], nil
+	}
+	if len(r) > 2 {
+		r0 := r[0]
+		r1, err := ResMerge(r[1:]...)
+		if err != nil {
+			return nil, err
+		}
+		return ResMerge(r0, r1)
+	}
+	// now we have r[0] and r[1] to merge here...
+	r0 := r[0]
+	r1 := r[1]
+	if err := AdaptCmp(r0, r1); err != nil {
+		return nil, err
+	}
+
+	res, err := r0.Merge(r1) // resource method of this interface
+	if err != nil {
+		return nil, err
+	}
+
+	// meta should have come over in the copy
+
+	if x, ok := res.(RefreshableRes); ok {
+		x0, ok0 := r0.(RefreshableRes)
+		x1, ok1 := r1.(RefreshableRes)
+		if !ok0 || !ok1 {
+			// programming error
+			panic("refresh interfaces are illogical")
+		}
+
+		x.SetRefresh(x0.Refresh() || x1.Refresh()) // true if either is!
+	}
+
+	// the other traits and metaparams can't be merged easily... so we don't
+	// merge them, and if they were present and differed, and weren't copied
+	// in the ResCopy method, then we should have errored above in AdaptCmp!
+
+	return res, nil
+}

engine/doc.go

@@ -0,0 +1,33 @@
+// Mgmt
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
+
+// Package engine represents the implementation of the resource engine that runs
+// the graph of resources in real-time. This package has the common imports that
+// most consumers use directly.
+package engine

engine/edge.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,7 +13,19 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package engine
 

engine/error.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,7 +13,19 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package engine
 
@@ -24,9 +36,9 @@ type Error string
 func (e Error) Error() string { return string(e) }
 
 const (
-	// ErrWatchExit represents an exit from the Watch loop via chan closure.
-	ErrWatchExit = Error("watch exit")
+	// ErrClosed means we couldn't complete a task because we had closed.
+	ErrClosed = Error("closed")
 
-	// ErrSignalExit represents an exit from the Watch loop via exit signal.
-	ErrSignalExit = Error("signal exit")
+	// ErrBackPoke means we're postponing due to a needed backpoke.
+	ErrBackPoke = Error("backpoke")
 )

engine/event/event.go

@@ -1,33 +0,0 @@
-// Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
-// Written by James Shubin <james@shubin.ca> and the project contributors
-//
-// This program is free software: you can redistribute it and/or modify
-// it under the terms of the GNU General Public License as published by
-// the Free Software Foundation, either version 3 of the License, or
-// (at your option) any later version.
-//
-// This program is distributed in the hope that it will be useful,
-// but WITHOUT ANY WARRANTY; without even the implied warranty of
-// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-// GNU General Public License for more details.
-//
-// You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
-
-// Package event provides some primitives that are used for message passing.
-package event
-
-//go:generate stringer -type=Kind -output=kind_stringer.go
-
-// Kind represents the type of event being passed.
-type Kind int
-
-// The different event kinds are used in different contexts.
-const (
-	EventNil Kind = iota
-	EventStart
-	EventPause
-	EventPoke
-	EventExit
-)

engine/fs.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,7 +13,19 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package engine
 
@@ -23,39 +35,64 @@ import (
 	"github.com/spf13/afero"
 )
 
-// from the ioutil package:
-// NopCloser(r io.Reader) io.ReadCloser // not implemented here
-// ReadAll(r io.Reader) ([]byte, error)
-// ReadDir(dirname string) ([]os.FileInfo, error)
-// ReadFile(filename string) ([]byte, error)
-// TempDir(dir, prefix string) (name string, err error)
-// TempFile(dir, prefix string) (f *os.File, err error) // slightly different here
-// WriteFile(filename string, data []byte, perm os.FileMode) error
-
-// Fs is an interface that represents this file system API that we support.
-// TODO: this should be in the gapi package or elsewhere.
+// Fs is an interface that represents the file system API that we support.
+// TODO: rename this to FS for consistency with the io/fs.FS naming scheme
 type Fs interface {
 	//fmt.Stringer // TODO: add this method?
-	afero.Fs     // TODO: why doesn't this interface exist in the os pkg?
+
+	// URI returns a unique string handle to access this filesystem.
 	URI() string // returns the URI for this file system
 
-	//DirExists(path string) (bool, error)
-	//Exists(path string) (bool, error)
-	//FileContainsAnyBytes(filename string, subslices [][]byte) (bool, error)
-	//FileContainsBytes(filename string, subslice []byte) (bool, error)
-	//FullBaseFsPath(basePathFs *BasePathFs, relativePath string) string
-	//GetTempDir(subPath string) string
-	//IsDir(path string) (bool, error)
-	//IsEmpty(path string) (bool, error)
-	//NeuterAccents(s string) string
-	//ReadAll(r io.Reader) ([]byte, error) // not needed
+	afero.Fs // TODO: why doesn't this interface exist in the os pkg?
+
+	// FS is the read-only filesystem interface from the io/fs.FS package.
+	//fs.FS // io/fs.FS
+
+	// ReadDir reads the named directory and returns a list of directory
+	// entries sorted by filename.
+	//
+	// This mimics the signature from io/fs.ReadDirFS and has the same docs.
+	//
+	// XXX: Not currently implemented because of legacy Afero.Fs above
+	//ReadDir(name string) ([]fs.DirEntry, error) // io/fs.ReadDirFS
+
+	// ReadFile reads the named file and returns its contents. A successful
+	// call returns a nil error, not io.EOF. (Because ReadFile reads the
+	// whole file, the expected EOF from the final Read is not treated as an
+	// error to be reported.)
+	//
+	// The caller is permitted to modify the returned byte slice. This
+	// method should return a copy of the underlying data.
+	//
+	// This mimics the signature from io/fs.ReadFileFS and has the same
+	// docs.
+	ReadFile(name string) ([]byte, error) // io/fs.ReadFileFS
+
+	// Stat returns a FileInfo describing the file. If there is an error, it
+	// should be of type *fs.PathError.
+	//
+	// This mimics the signature from io/fs.StatFS and has the same docs.
+	//
+	// XXX: Not currently implemented because of legacy Afero.Fs above
+	//Stat(name string) (FileInfo, error) // io/fs.StatFS
+
+	// afero.Fs versions:
+
 	ReadDir(dirname string) ([]os.FileInfo, error)
-	ReadFile(filename string) ([]byte, error)
-	//SafeWriteReader(path string, r io.Reader) (err error)
-	TempDir(dir, prefix string) (name string, err error)
-	TempFile(dir, prefix string) (f afero.File, err error) // slightly different from upstream
-	//UnicodeSanitize(s string) string
-	//Walk(root string, walkFn filepath.WalkFunc) error
-	WriteFile(filename string, data []byte, perm os.FileMode) error
-	//WriteReader(path string, r io.Reader) (err error)
+}
+
+// WriteableFS is our internal filesystem interface for filesystems we write to.
+// It can wrap whatever implementations we want.
+type WriteableFS interface {
+	Fs
+
+	// WriteFile writes data to the named file, creating it if necessary. If
+	// the file does not exist, WriteFile creates it with permissions perm
+	// (before umask); otherwise WriteFile truncates it before writing,
+	// without changing permissions. Since Writefile requires multiple
+	// system calls to complete, a failure mid-operation can leave the file
+	// in a partially written state.
+	//
+	// This mimics the internal os.WriteFile function and has the same docs.
+	WriteFile(name string, data []byte, perm os.FileMode) error
 }

engine/graph/actions.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,22 +13,34 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package graph
 
 import (
+	"context"
 	"fmt"
 	"strings"
 	"sync"
 	"time"
 
 	"github.com/purpleidea/mgmt/engine"
-	"github.com/purpleidea/mgmt/engine/event"
+	engineUtil "github.com/purpleidea/mgmt/engine/util"
 	"github.com/purpleidea/mgmt/pgraph"
+	"github.com/purpleidea/mgmt/util/errwrap"
 
-	//multierr "github.com/hashicorp/go-multierror"
-	errwrap "github.com/pkg/errors"
 	"golang.org/x/time/rate"
 )
 
@@ -41,14 +53,18 @@ func (obj *Engine) OKTimestamp(vertex pgraph.Vertex) bool {
 // be bad.
 func (obj *Engine) BadTimestamps(vertex pgraph.Vertex) []pgraph.Vertex {
 	vs := []pgraph.Vertex{}
-	ts := obj.state[vertex].timestamp
+	obj.state[vertex].mutex.RLock()   // concurrent read start
+	ts := obj.state[vertex].timestamp // race
+	obj.state[vertex].mutex.RUnlock() // concurrent read end
 	// these are all the vertices pointing TO vertex, eg: ??? -> vertex
 	for _, v := range obj.graph.IncomingGraphVertices(vertex) {
 		// If the vertex has a greater timestamp than any prerequisite,
 		// then we can't run right now. If they're equal (eg: initially
 		// with a value of 0) then we also can't run because we should
 		// let our pre-requisites go first.
-		t := obj.state[v].timestamp
+		obj.state[v].mutex.RLock()   // concurrent read start
+		t := obj.state[v].timestamp  // race
+		obj.state[v].mutex.RUnlock() // concurrent read end
 		if obj.Debug {
 			obj.Logf("OKTimestamp: %d >= %d (%s): !%t", ts, t, v.String(), ts >= t)
 		}
@@ -61,33 +77,33 @@ func (obj *Engine) BadTimestamps(vertex pgraph.Vertex) []pgraph.Vertex {
 }
 
 // Process is the primary function to execute a particular vertex in the graph.
-func (obj *Engine) Process(vertex pgraph.Vertex) error {
+func (obj *Engine) Process(ctx context.Context, vertex pgraph.Vertex) error {
 	res, isRes := vertex.(engine.Res)
 	if !isRes {
 		return fmt.Errorf("vertex is not a Res")
 	}
 
-	// Engine Guarantee: Do not allow CheckApply to run while we are paused.
-	// This makes the resource able to know that synchronous channel sending
-	// to the main loop select in Watch from within CheckApply, will succeed
-	// without blocking because the resource went into a paused state. If we
-	// are using the Poll metaparam, then Watch will (of course) not be run.
-	// FIXME: should this lock be here, or wrapped right around CheckApply ?
-	obj.state[vertex].eventsLock.Lock() // this lock is taken within Event()
-	defer obj.state[vertex].eventsLock.Unlock()
-
 	// backpoke! (can be async)
 	if vs := obj.BadTimestamps(vertex); len(vs) > 0 {
 		// back poke in parallel (sync b/c of waitgroup)
+		wg := &sync.WaitGroup{}
 		for _, v := range obj.graph.IncomingGraphVertices(vertex) {
 			if !pgraph.VertexContains(v, vs) { // only poke what's needed
 				continue
 			}
 
-			go obj.state[v].Poke() // async
+			// doesn't really need to be in parallel, but we can...
+			wg.Add(1)
+			go func(vv pgraph.Vertex) {
+				defer wg.Done()
+				obj.state[vv].Poke() // async
+			}(v)
 
 		}
-		return nil // can't continue until timestamp is in sequence
+		wg.Wait()
+
+		// can't continue until timestamp is in sequence, defer for now
+		return engine.ErrBackPoke
 	}
 
 	// semaphores!
@@ -112,20 +128,41 @@ func (obj *Engine) Process(vertex pgraph.Vertex) error {
 
 	// sendrecv!
 	// connect any senders to receivers and detect if values changed
+	// this actually checks and sends into resource trees recursively...
 	if res, ok := vertex.(engine.RecvableRes); ok {
-		if updated, err := obj.SendRecv(res); err != nil {
+		if obj.Debug {
+			obj.Logf("SendRecv: %s", res) // receiving here
+		}
+		if updated, err := SendRecv(res, nil); err != nil {
 			return errwrap.Wrapf(err, "could not SendRecv")
 		} else if len(updated) > 0 {
-			for _, changed := range updated {
-				if changed { // at least one was updated
-					// invalidate cache, mark as dirty
-					obj.state[vertex].isStateOK = false
-					break
+			//for _, s := range graph.UpdatedStrings(updated) {
+			//	obj.Logf("SendRecv: %s", s)
+			//}
+			for r, m := range updated { // map[engine.RecvableRes]map[string]*engine.Send
+				v, ok := r.(pgraph.Vertex)
+				if !ok {
+					continue
+				}
+				_, stateExists := obj.state[v] // autogrouped children probably don't have a state
+				if !stateExists {
+					continue
+				}
+				for s, send := range m {
+					if !send.Changed {
+						continue
+					}
+					obj.Logf("Send/Recv: %v.%s -> %v.%s", send.Res, send.Key, r, s)
+					// if send.Changed == true, at least one was updated
+					// invalidate cache, mark as dirty
+					obj.state[v].setDirty()
+					//break // we might have more vertices now
+				}
+
+				// re-validate after we change any values
+				if err := engine.Validate(r); err != nil {
+					return errwrap.Wrapf(err, "failed Validate after SendRecv")
 				}
-			}
-			// re-validate after we change any values
-			if err := engine.Validate(res); err != nil {
-				return errwrap.Wrapf(err, "failed Validate after SendRecv")
 			}
 		}
 	}
@@ -147,18 +184,18 @@ func (obj *Engine) Process(vertex pgraph.Vertex) error {
 	// Check cached state, to skip CheckApply, but can't skip if refreshing!
 	// If the resource doesn't implement refresh, skip the refresh test.
 	// FIXME: if desired, check that we pass through refresh notifications!
-	if (!refresh || !isRefreshableRes) && obj.state[vertex].isStateOK {
+	if (!refresh || !isRefreshableRes) && obj.state[vertex].isStateOK.Load() { // mutex RLock/RUnlock
 		checkOK, err = true, nil
 
 	} else if noop && (refresh && isRefreshableRes) { // had a refresh to do w/ noop!
 		checkOK, err = false, nil // therefore the state is wrong
 
-		// run the CheckApply!
 	} else {
+		// run the CheckApply!
 		obj.Logf("%s: CheckApply(%t)", res, !noop)
 		// if this fails, don't UpdateTimestamp()
-		checkOK, err = res.CheckApply(!noop)
-		obj.Logf("%s: CheckApply(%t): Return(%t, %+v)", res, !noop, checkOK, err)
+		checkOK, err = res.CheckApply(ctx, !noop)
+		obj.Logf("%s: CheckApply(%t): Return(%t, %s)", res, !noop, checkOK, engineUtil.CleanError(err))
 	}
 
 	if checkOK && err != nil { // should never return this way
@@ -174,7 +211,10 @@ func (obj *Engine) Process(vertex pgraph.Vertex) error {
 
 	// if CheckApply ran without noop and without error, state should be good
 	if !noop && err == nil { // aka !noop || checkOK
-		obj.state[vertex].isStateOK = true // reset
+		obj.state[vertex].tuid.StartTimer()
+		//obj.state[vertex].mutex.Lock()
+		obj.state[vertex].isStateOK.Store(true) // reset
+		//obj.state[vertex].mutex.Unlock()
 		if refresh {
 			obj.SetUpstreamRefresh(vertex, false) // refresh happened, clear the request
 			if isRefreshableRes {
@@ -211,7 +251,9 @@ func (obj *Engine) Process(vertex pgraph.Vertex) error {
 		wg := &sync.WaitGroup{}
 		// update this timestamp *before* we poke or the poked
 		// nodes might fail due to having a too old timestamp!
-		obj.state[vertex].timestamp = time.Now().UnixNano() // update timestamp
+		obj.state[vertex].mutex.Lock()                      // concurrent write start
+		obj.state[vertex].timestamp = time.Now().UnixNano() // update timestamp (race)
+		obj.state[vertex].mutex.Unlock()                    // concurrent write end
 		for _, v := range obj.graph.OutgoingGraphVertices(vertex) {
 			if !obj.OKTimestamp(v) {
 				// there is at least another one that will poke this...
@@ -242,26 +284,68 @@ func (obj *Engine) Process(vertex pgraph.Vertex) error {
 
 // Worker is the common run frontend of the vertex. It handles all of the retry
 // and retry delay common code, and ultimately returns the final status of this
-// vertex execution.
+// vertex execution. This function cannot be "re-run" for the same vertex. The
+// retry mechanism stuff happens inside of this. To actually "re-run" you need
+// to remove the vertex and build a new one. The engine guarantees that we do
+// not allow CheckApply to run while we are paused. That is enforced here.
 func (obj *Engine) Worker(vertex pgraph.Vertex) error {
 	res, isRes := vertex.(engine.Res)
 	if !isRes {
 		return fmt.Errorf("vertex is not a resource")
 	}
 
-	defer close(obj.state[vertex].stopped) // done signal
+	// bonus safety check
+	if res.MetaParams().Burst == 0 && !(res.MetaParams().Limit == rate.Inf) { // blocked
+		return fmt.Errorf("permanently limited (rate != Inf, burst = 0)")
+	}
+
+	// initialize or reinitialize the meta state for this resource uid
+	obj.mlock.Lock()
+	if _, exists := obj.metas[engine.PtrUID(res)]; !exists || res.MetaParams().Reset {
+		obj.metas[engine.PtrUID(res)] = &engine.MetaState{
+			CheckApplyRetry: res.MetaParams().Retry, // lookup the retry value
+		}
+	}
+	metas := obj.metas[engine.PtrUID(res)] // handle
+	obj.mlock.Unlock()
+
+	//defer close(obj.state[vertex].stopped) // done signal
 
 	obj.state[vertex].cuid = obj.Converger.Register()
+	obj.state[vertex].tuid = obj.Converger.Register()
 	// must wait for all users of the cuid to finish *before* we unregister!
 	// as a result, this defer happens *before* the below wait group Wait...
 	defer obj.state[vertex].cuid.Unregister()
+	defer obj.state[vertex].tuid.Unregister()
 
 	defer obj.state[vertex].wg.Wait() // this Worker is the last to exit!
 
 	obj.state[vertex].wg.Add(1)
 	go func() {
 		defer obj.state[vertex].wg.Done()
-		defer close(obj.state[vertex].outputChan) // we close this on behalf of res
+		defer close(obj.state[vertex].eventsChan) // we close this on behalf of res
+
+		// This is a close reverse-multiplexer. If any of the channels
+		// close, then it will cause the doneCtx to cancel. That way,
+		// multiple different folks can send a close signal, without
+		// every worrying about duplicate channel close panics.
+		obj.state[vertex].wg.Add(1)
+		go func() {
+			defer obj.state[vertex].wg.Done()
+
+			// reverse-multiplexer: any close, causes *the* close!
+			select {
+			case <-obj.state[vertex].processDone:
+			case <-obj.state[vertex].watchDone:
+			case <-obj.state[vertex].limitDone:
+			case <-obj.state[vertex].retryDone:
+			case <-obj.state[vertex].removeDone:
+			case <-obj.state[vertex].eventsDone:
+			}
+
+			// the main "done" signal gets activated here!
+			obj.state[vertex].doneCtxCancel() // cancels doneCtx
+		}()
 
 		var err error
 		var retry = res.MetaParams().Retry // lookup the retry value
@@ -279,13 +363,8 @@ func (obj *Engine) Worker(vertex pgraph.Vertex) error {
 						case <-timer.C: // the wait is over
 							return errDelayExpired // special
 
-						case event, ok := <-obj.state[vertex].init.Events:
-							if !ok {
-								return nil
-							}
-							if err := obj.state[vertex].init.Read(event); err != nil {
-								return err
-							}
+						case <-obj.state[vertex].doneCtx.Done():
+							return nil
 						}
 					}
 				}()
@@ -295,180 +374,312 @@ func (obj *Engine) Worker(vertex pgraph.Vertex) error {
 				}
 			} else if interval := res.MetaParams().Poll; interval > 0 { // poll instead of watching :(
 				obj.state[vertex].cuid.StartTimer()
-				err = obj.state[vertex].poll(interval)
+				err = obj.state[vertex].poll(obj.state[vertex].doneCtx, interval)
 				obj.state[vertex].cuid.StopTimer() // clean up nicely
 			} else {
 				obj.state[vertex].cuid.StartTimer()
-				obj.Logf("Watch(%s)", vertex)
-				err = res.Watch() // run the watch normally
-				obj.Logf("Watch(%s): Exited(%+v)", vertex, err)
+				if obj.Debug {
+					obj.Logf("%s: Watch...", vertex)
+				}
+				err = res.Watch(obj.state[vertex].doneCtx) // run the watch normally
+				if obj.Debug {
+					if s := engineUtil.CleanError(err); err != nil {
+						obj.Logf("%s: Watch Error: %s", vertex, s)
+					} else {
+						obj.Logf("%s: Watch Exited...", vertex)
+					}
+				}
 				obj.state[vertex].cuid.StopTimer() // clean up nicely
 			}
-			if err == nil || err == engine.ErrWatchExit || err == engine.ErrSignalExit {
+			if err == nil { // || err == engine.ErrClosed
 				return // exited cleanly, we're done
 			}
 			// we've got an error...
 			delay = res.MetaParams().Delay
 
 			if retry < 0 { // infinite retries
-				obj.state[vertex].reset()
 				continue
 			}
 			if retry > 0 { // don't decrement past 0
 				retry--
 				obj.state[vertex].init.Logf("retrying Watch after %.4f seconds (%d left)", float64(delay)/1000, retry)
-				obj.state[vertex].reset()
 				continue
 			}
 			//if retry == 0 { // optional
 			//	err = errwrap.Wrapf(err, "permanent watch error")
 			//}
 			break // break out of this and send the error
-		}
+		} // for retry loop
+
 		// this section sends an error...
 		// If the CheckApply loop exits and THEN the Watch fails with an
 		// error, then we'd be stuck here if exit signal didn't unblock!
 		select {
-		case obj.state[vertex].outputChan <- errwrap.Wrapf(err, "watch failed"):
+		case obj.state[vertex].eventsChan <- errwrap.Wrapf(err, "watch failed"):
 			// send
-		case <-obj.state[vertex].exit.Signal():
-			// pass
 		}
 	}()
 
-	// bonus safety check
-	if res.MetaParams().Burst == 0 && !(res.MetaParams().Limit == rate.Inf) { // blocked
-		return fmt.Errorf("permanently limited (rate != Inf, burst = 0)")
-	}
-	var limiter = rate.NewLimiter(res.MetaParams().Limit, res.MetaParams().Burst)
-	// It is important that we shutdown the Watch loop if this exits.
-	// Example, if Process errors permanently, we should ask Watch to exit.
-	defer obj.state[vertex].Event(event.EventExit) // signal an exit
-	for {
+	// If this exits cleanly, we must unblock the reverse-multiplexer.
+	// I think this additional close is unnecessary, but it's not harmful.
+	defer close(obj.state[vertex].eventsDone) // causes doneCtx to cancel
+	limiter := rate.NewLimiter(res.MetaParams().Limit, res.MetaParams().Burst)
+	var reserv *rate.Reservation
+	var reterr error
+	var failed bool // has Process permanently failed?
+	var closed bool // has the resumeSignal channel closed?
+Loop:
+	for { // process loop
+		// This is the main select where things happen and where we exit
+		// from. It's similar to the two "satellite" select's which we
+		// might spend some time in if we're retrying or rate limiting.
+		// This select is also the main event receiver and is also the
+		// only place where we read from the poke channel.
 		select {
-		case err, ok := <-obj.state[vertex].outputChan: // read from watch channel
+		case err, ok := <-obj.state[vertex].eventsChan: // read from watch channel
 			if !ok {
-				return nil
+				return reterr // we only return when chan closes
 			}
+			// If the Watch method exits with an error, then this
+			// channel will get that error propagated to it, which
+			// we then save so we can return it to the caller of us.
 			if err != nil {
-				return err // permanent failure
+				failed = true
+				close(obj.state[vertex].watchDone)   // causes doneCtx to cancel
+				reterr = errwrap.Append(reterr, err) // permanent failure
+				continue
 			}
+			if obj.Debug {
+				obj.Logf("event received")
+			}
+			reserv = limiter.ReserveN(time.Now(), 1) // one event
+			// reserv.OK() seems to always be true here!
 
-			// safe to go run the process...
-		case <-obj.state[vertex].exit.Signal(): // TODO: is this needed?
-			return nil
+		case _, ok := <-obj.state[vertex].pokeChan: // read from buffered poke channel
+			if !ok { // we never close it
+				panic("unexpected close of poke channel")
+			}
+			if obj.Debug {
+				obj.Logf("poke received")
+			}
+			reserv = nil // we didn't receive a real event here...
+
+		case _, ok := <-obj.state[vertex].pauseSignal: // one message
+			if !ok {
+				obj.state[vertex].pauseSignal = nil
+				continue // this is not a new pause message
+			}
+			// NOTE: If we allowed a doneCtx below to let us out
+			// of the resumeSignal wait, then we could loop around
+			// and run this again, causing a panic. Instead of this
+			// being made safe with a sync.Once, we instead run a
+			// close() call inside of the vertexRemoveFn function,
+			// which should unblock resumeSignal so we can shutdown.
+
+			// we are paused now, and waiting for resume or exit...
+			select {
+			case _, ok := <-obj.state[vertex].resumeSignal: // channel closes
+				if !ok {
+					closed = true
+				}
+				// resumed!
+				// pass through to allow a Process to try to run
+				// TODO: consider adding this fast pause here...
+				//if obj.fastPause {
+				//	obj.Logf("fast pausing on resume")
+				//	continue
+				//}
+			}
 		}
 
-		now := time.Now()
-		r := limiter.ReserveN(now, 1) // one event
-		// r.OK() seems to always be true here!
-		d := r.DelayFrom(now)
-		if d > 0 { // delay
-			obj.state[vertex].init.Logf("limited (rate: %v/sec, burst: %d, next: %v)", res.MetaParams().Limit, res.MetaParams().Burst, d)
-			var count int
+		// drop redundant pokes
+		for len(obj.state[vertex].pokeChan) > 0 {
+			select {
+			case <-obj.state[vertex].pokeChan:
+			default:
+				// race, someone else read one!
+			}
+		}
+
+		// don't Process anymore if we've already failed or shutdown...
+		if failed || closed {
+			continue Loop
+		}
+
+		// limit delay
+		d := time.Duration(0)
+		if reserv != nil {
+			d = reserv.DelayFrom(time.Now())
+		}
+		if reserv != nil && d > 0 { // delay
+			obj.state[vertex].init.Logf("limited (rate: %v/sec, burst: %d, next: %dms)", res.MetaParams().Limit, res.MetaParams().Burst, d/time.Millisecond)
 			timer := time.NewTimer(time.Duration(d) * time.Millisecond)
 		LimitWait:
 			for {
+				// This "satellite" select doesn't need a poke
+				// channel because we're already in "event
+				// received" mode, and poke doesn't block.
 				select {
 				case <-timer.C: // the wait is over
 					break LimitWait
 
 				// consume other events while we're waiting...
-				case e, ok := <-obj.state[vertex].outputChan: // read from watch channel
+				case e, ok := <-obj.state[vertex].eventsChan: // read from watch channel
 					if !ok {
-						// FIXME: is this logic correct?
-						if count == 0 {
-							return nil
-						}
-						// loop, because we have
-						// the previous event to
-						// run process on first!
-						continue
+						return reterr // we only return when chan closes
 					}
 					if e != nil {
-						return e // permanent failure
+						failed = true
+						close(obj.state[vertex].limitDone) // causes doneCtx to cancel
+						reterr = errwrap.Append(reterr, e) // permanent failure
+						break LimitWait
 					}
-					count++                         // count the events...
+					if obj.Debug {
+						obj.Logf("event received in limit")
+					}
+					// TODO: does this get added in properly?
 					limiter.ReserveN(time.Now(), 1) // one event
+
+				// this pause/resume block is the same as the upper main one
+				case _, ok := <-obj.state[vertex].pauseSignal:
+					if !ok {
+						obj.state[vertex].pauseSignal = nil
+						break LimitWait
+					}
+					select {
+					case _, ok := <-obj.state[vertex].resumeSignal: // channel closes
+						if !ok {
+							closed = true
+						}
+						// resumed!
+					}
 				}
 			}
 			timer.Stop() // it's nice to cleanup
 			obj.state[vertex].init.Logf("rate limiting expired!")
 		}
+		// don't Process anymore if we've already failed or shutdown...
+		if failed || closed {
+			continue Loop
+		}
+		// end of limit delay
 
+		// retry...
 		var err error
-		var retry = res.MetaParams().Retry // lookup the retry value
+		//var retry = res.MetaParams().Retry // lookup the retry value
 		var delay uint64
-	Loop:
+	RetryLoop:
 		for { // retry loop
 			if delay > 0 {
-				var count int
 				timer := time.NewTimer(time.Duration(delay) * time.Millisecond)
 			RetryWait:
 				for {
+					// This "satellite" select doesn't need
+					// a poke channel because we're already
+					// in "event received" mode, and poke
+					// doesn't block.
 					select {
 					case <-timer.C: // the wait is over
 						break RetryWait
 
 					// consume other events while we're waiting...
-					case e, ok := <-obj.state[vertex].outputChan: // read from watch channel
+					case e, ok := <-obj.state[vertex].eventsChan: // read from watch channel
 						if !ok {
-							// FIXME: is this logic correct?
-							if count == 0 {
-								// last process error
-								return err
-							}
-							// loop, because we have
-							// the previous event to
-							// run process on first!
-							continue
+							return reterr // we only return when chan closes
 						}
 						if e != nil {
-							return e // permanent failure
+							failed = true
+							close(obj.state[vertex].retryDone) // causes doneCtx to cancel
+							reterr = errwrap.Append(reterr, e) // permanent failure
+							break RetryWait
 						}
-						count++                         // count the events...
+						if obj.Debug {
+							obj.Logf("event received in retry")
+						}
+						// TODO: does this get added in properly?
 						limiter.ReserveN(time.Now(), 1) // one event
+
+					// this pause/resume block is the same as the upper main one
+					case _, ok := <-obj.state[vertex].pauseSignal:
+						if !ok {
+							obj.state[vertex].pauseSignal = nil
+							break RetryWait
+						}
+						select {
+						case _, ok := <-obj.state[vertex].resumeSignal: // channel closes
+							if !ok {
+								closed = true
+							}
+							// resumed!
+						}
 					}
 				}
 				timer.Stop() // it's nice to cleanup
 				delay = 0    // reset
 				obj.state[vertex].init.Logf("the CheckApply delay expired!")
 			}
+			// don't Process anymore if we've already failed or shutdown...
+			if failed || closed {
+				continue Loop
+			}
 
 			if obj.Debug {
 				obj.Logf("Process(%s)", vertex)
 			}
-			err = obj.Process(vertex)
-			if obj.Debug {
-				obj.Logf("Process(%s): Return(%+v)", vertex, err)
+			backPoke := false
+			err = obj.Process(obj.state[vertex].doneCtx, vertex)
+			if err == engine.ErrBackPoke {
+				backPoke = true
+				err = nil // for future code safety
 			}
-			if err == nil {
-				break Loop
+			if obj.Debug && backPoke {
+				obj.Logf("Process(%s): BackPoke!", vertex)
+			}
+			if obj.Debug && !backPoke {
+				obj.Logf("Process(%s): Return(%s)", vertex, engineUtil.CleanError(err))
+			}
+			if err == nil && !backPoke && res.MetaParams().RetryReset { // reset it on success!
+				metas.CheckApplyRetry = res.MetaParams().Retry // lookup the retry value
+			}
+			if err == nil || backPoke {
+				break RetryLoop
 			}
 			// we've got an error...
 			delay = res.MetaParams().Delay
 
-			if retry < 0 { // infinite retries
+			if metas.CheckApplyRetry < 0 { // infinite retries
 				continue
 			}
-			if retry > 0 { // don't decrement past 0
-				retry--
-				obj.state[vertex].init.Logf("retrying CheckApply after %.4f seconds (%d left)", float64(delay)/1000, retry)
+			if metas.CheckApplyRetry > 0 { // don't decrement past 0
+				metas.CheckApplyRetry--
+				obj.state[vertex].init.Logf(
+					"retrying CheckApply after %.4f seconds (%d left)",
+					float64(delay)/1000,
+					metas.CheckApplyRetry,
+				)
 				continue
 			}
-			//if retry == 0 { // optional
+			//if metas.CheckApplyRetry == 0 { // optional
 			//	err = errwrap.Wrapf(err, "permanent process error")
 			//}
 
-			// If this exits, defer calls Event(event.EventExit),
-			// which will cause the Watch loop to shutdown. Also,
-			// if the Watch loop shuts down, that will cause this
-			// Process loop to shut down. Also the graph sync can
-			// run an Event(event.EventExit) which causes this to
-			// shutdown as well. Lastly, it is possible that more
-			// that one of these scenarios happens simultaneously.
-			return err
-		}
-	}
+			// It is important that we shutdown the Watch loop if
+			// this dies. If Process fails permanently, we ask it
+			// to exit right here... (It happens when we loop...)
+			failed = true
+			close(obj.state[vertex].processDone) // causes doneCtx to cancel
+			reterr = errwrap.Append(reterr, err) // permanent failure
+			continue
+
+		} // retry loop
+
+		// When this Process loop exits, it's because something has
+		// caused Watch() to shutdown (even if it's our permanent
+		// failure from Process), which caused this channel to close.
+		// On or more exit signals are possible, and more than one can
+		// happen simultaneously.
+
+	} // process loop
+
 	//return nil // unreachable
 }

engine/graph/autoedge.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,7 +13,19 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package graph
 

engine/graph/autoedge/autoedge.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,7 +13,19 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package autoedge
 
@@ -22,9 +34,7 @@ import (
 
 	"github.com/purpleidea/mgmt/engine"
 	"github.com/purpleidea/mgmt/pgraph"
-
-	multierr "github.com/hashicorp/go-multierror"
-	errwrap "github.com/pkg/errors"
+	"github.com/purpleidea/mgmt/util/errwrap"
 )
 
 // AutoEdge adds the automatic edges to the graph.
@@ -49,7 +59,7 @@ func AutoEdge(graph *pgraph.Graph, debug bool, logf func(format string, v ...int
 	for _, res := range sorted { // for each vertexes autoedges
 		autoEdgeObj, e := res.AutoEdges()
 		if e != nil {
-			err = multierr.Append(err, e) // collect all errors
+			err = errwrap.Append(err, e) // collect all errors
 			continue
 		}
 		if autoEdgeObj == nil {
@@ -91,6 +101,9 @@ func AutoEdge(graph *pgraph.Graph, debug bool, logf func(format string, v ...int
 			}
 		}
 	}
+
+	// It would be great to ensure we didn't add any graph cycles here, but
+	// instead of checking now, we'll move the check into the main loop.
 	return nil
 }
 

engine/graph/autogroup.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,7 +13,19 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package graph
 
@@ -24,8 +36,7 @@ import (
 	"github.com/purpleidea/mgmt/engine/graph/autogroup"
 	"github.com/purpleidea/mgmt/pgraph"
 	"github.com/purpleidea/mgmt/util"
-
-	errwrap "github.com/pkg/errors"
+	"github.com/purpleidea/mgmt/util/errwrap"
 )
 
 // AutoGroup runs the auto grouping on the loaded graph.
@@ -75,10 +86,10 @@ func (obj *wrappedGrouper) VertexCmp(v1, v2 pgraph.Vertex) error {
 		return fmt.Errorf("v2 is not a GroupableRes")
 	}
 
-	if r1.Kind() != r2.Kind() { // we must group similar kinds
-		// TODO: maybe future resources won't need this limitation?
-		return fmt.Errorf("the two resources aren't the same kind")
-	}
+	// Some resources of different kinds can now group together!
+	//if r1.Kind() != r2.Kind() { // we must group similar kinds
+	//	return fmt.Errorf("the two resources aren't the same kind")
+	//}
 	// someone doesn't want to group!
 	if r1.AutoGroupMeta().Disabled || r2.AutoGroupMeta().Disabled {
 		return fmt.Errorf("one of the autogroup flags is false")
@@ -110,6 +121,7 @@ func (obj *wrappedGrouper) VertexMerge(v1, v2 pgraph.Vertex) (v pgraph.Vertex, e
 	if err = r1.GroupRes(r2); err != nil { // GroupRes skips stupid groupings
 		return // return early on error
 	}
+	r2.SetParent(r1) // store who my parent is
 
 	// merging two resources into one should yield the sum of their semas
 	if semas := r2.MetaParams().Sema; len(semas) > 0 {

engine/graph/autogroup/autogroup.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,7 +13,19 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package autogroup
 
@@ -22,8 +34,7 @@ import (
 
 	"github.com/purpleidea/mgmt/engine"
 	"github.com/purpleidea/mgmt/pgraph"
-
-	errwrap "github.com/pkg/errors"
+	"github.com/purpleidea/mgmt/util/errwrap"
 )
 
 // AutoGroup is the mechanical auto group "runner" that runs the interface spec.
@@ -55,7 +66,7 @@ func AutoGroup(ag engine.AutoGrouper, g *pgraph.Graph, debug bool, logf func(for
 			logf("!VertexMerge for: %s into: %s", wStr, vStr)
 
 		} else { // success!
-			logf("success for: %s into: %s", wStr, vStr)
+			logf("%s into %s", wStr, vStr)
 			merged = true // woo
 		}
 
@@ -67,5 +78,8 @@ func AutoGroup(ag engine.AutoGrouper, g *pgraph.Graph, debug bool, logf func(for
 		}
 	}
 
+	// It would be great to ensure we didn't add any graph cycles here, but
+	// instead of checking now, we'll move the check into the main loop.
+
 	return nil
 }

engine/graph/autogroup/autogroup_test.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,13 +13,26 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
-// +build !root
+//go:build !root
 
 package autogroup
 
 import (
+	"context"
 	"fmt"
 	"reflect"
 	"sort"
@@ -31,8 +44,7 @@ import (
 	"github.com/purpleidea/mgmt/engine/traits"
 	"github.com/purpleidea/mgmt/pgraph"
 	"github.com/purpleidea/mgmt/util"
-
-	errwrap "github.com/pkg/errors"
+	"github.com/purpleidea/mgmt/util/errwrap"
 )
 
 func init() {
@@ -63,15 +75,15 @@ func (obj *NoopResTest) Init(init *engine.Init) error {
 	return nil
 }
 
-func (obj *NoopResTest) Close() error {
+func (obj *NoopResTest) Cleanup() error {
 	return nil
 }
 
-func (obj *NoopResTest) Watch() error {
+func (obj *NoopResTest) Watch(context.Context) error {
 	return nil // not needed
 }
 
-func (obj *NoopResTest) CheckApply(apply bool) (checkOK bool, err error) {
+func (obj *NoopResTest) CheckApply(ctx context.Context, apply bool) (checkOK bool, err error) {
 	return true, nil // state is always okay
 }
 
@@ -190,6 +202,8 @@ func (obj *testGrouper) VertexMerge(v1, v2 pgraph.Vertex) (v pgraph.Vertex, err
 	if err := r1.GroupRes(r2); err != nil { // group them first
 		return nil, err
 	}
+	r2.SetParent(r1) // store who my parent is
+
 	// HACK: update the name so it matches full list of self+grouped
 	res := v1.(engine.GroupableRes)
 	names := strings.Split(res.Name(), ",") // load in stored names
@@ -596,10 +610,12 @@ func TestPgraphGrouping11(t *testing.T) {
 	runGraphCmp(t, g1, g2)
 }
 
+/*
 // simple merge 1
 // a1   a2         a1,a2
 //   \ /     >>>     |     (arrows point downwards)
 //    b              b
+*/
 func TestPgraphGrouping12(t *testing.T) {
 	g1, _ := pgraph.NewGraph("g1") // original graph
 	{
@@ -621,10 +637,12 @@ func TestPgraphGrouping12(t *testing.T) {
 	runGraphCmp(t, g1, g2)
 }
 
+/*
 // simple merge 2
 //    b              b
 //   / \     >>>     |     (arrows point downwards)
 // a1   a2         a1,a2
+*/
 func TestPgraphGrouping13(t *testing.T) {
 	g1, _ := pgraph.NewGraph("g1") // original graph
 	{
@@ -646,10 +664,12 @@ func TestPgraphGrouping13(t *testing.T) {
 	runGraphCmp(t, g1, g2)
 }
 
+/*
 // triple merge
 // a1 a2  a3         a1,a2,a3
 //   \ | /     >>>       |      (arrows point downwards)
 //     b                 b
+*/
 func TestPgraphGrouping14(t *testing.T) {
 	g1, _ := pgraph.NewGraph("g1") // original graph
 	{
@@ -674,12 +694,14 @@ func TestPgraphGrouping14(t *testing.T) {
 	runGraphCmp(t, g1, g2)
 }
 
+/*
 // chain merge
 //    a1             a1
 //   /  \             |
 // b1    b2   >>>   b1,b2   (arrows point downwards)
 //   \  /             |
 //    c1             c1
+*/
 func TestPgraphGrouping15(t *testing.T) {
 	g1, _ := pgraph.NewGraph("g1") // original graph
 	{
@@ -709,6 +731,7 @@ func TestPgraphGrouping15(t *testing.T) {
 	runGraphCmp(t, g1, g2)
 }
 
+/*
 // re-attach 1 (outer)
 // technically the second possibility is valid too, depending on which order we
 // merge edges in, and if we don't filter out any unnecessary edges afterwards!
@@ -717,6 +740,7 @@ func TestPgraphGrouping15(t *testing.T) {
 // b1  /      >>>    b1     OR    b1  /   (arrows point downwards)
 //  | /               |            | /
 // c1                c1           c1
+*/
 func TestPgraphGrouping16(t *testing.T) {
 	g1, _ := pgraph.NewGraph("g1") // original graph
 	{
@@ -731,25 +755,40 @@ func TestPgraphGrouping16(t *testing.T) {
 		g1.AddEdge(b1, c1, e2)
 		g1.AddEdge(a2, c1, e3)
 	}
-	g2, _ := pgraph.NewGraph("g2") // expected result
+	//g2, _ := pgraph.NewGraph("g2") // expected result
+	//{
+	//	a := NewNoopResTest("a1,a2")
+	//	b1 := NewNoopResTest("b1")
+	//	c1 := NewNoopResTest("c1")
+	//	e1 := NE("e1,e3")
+	//	e2 := NE("e2,e3") // e3 gets "merged through" to BOTH edges!
+	//	g2.AddEdge(a, b1, e1)
+	//	g2.AddEdge(b1, c1, e2)
+	//}
+	//runGraphCmp(t, g1, g2)
+	g3, _ := pgraph.NewGraph("g3") // alternative expected result
 	{
 		a := NewNoopResTest("a1,a2")
 		b1 := NewNoopResTest("b1")
 		c1 := NewNoopResTest("c1")
-		e1 := NE("e1,e3")
-		e2 := NE("e2,e3") // e3 gets "merged through" to BOTH edges!
-		g2.AddEdge(a, b1, e1)
-		g2.AddEdge(b1, c1, e2)
+		e1 := NE("e1")
+		e2 := NE("e2")
+		e3 := NE("e3")
+		g3.AddEdge(a, b1, e1)
+		g3.AddEdge(b1, c1, e2)
+		g3.AddEdge(a, c1, e3)
 	}
-	runGraphCmp(t, g1, g2)
+	runGraphCmp(t, g1, g3)
 }
 
+/*
 // re-attach 2 (inner)
 // a1    b2          a1
 //  |   /             |
 // b1  /      >>>   b1,b2   (arrows point downwards)
 //  | /               |
 // c1                c1
+*/
 func TestPgraphGrouping17(t *testing.T) {
 	g1, _ := pgraph.NewGraph("g1") // original graph
 	{
@@ -777,13 +816,16 @@ func TestPgraphGrouping17(t *testing.T) {
 	runGraphCmp(t, g1, g2)
 }
 
+/*
 // re-attach 3 (double)
 // similar to "re-attach 1", technically there is a second possibility for this
-// a2   a1    b2         a1,a2
-//   \   |   /             |
-//    \ b1  /      >>>   b1,b2   (arrows point downwards)
-//     \ | /               |
-//      c1                c1
+// TODO: verify this second possibility manually
+// a2   a1    b2         a1,a2        a1,a2
+//   \   |   /             |            |   \
+//    \ b1  /      >>>   b1,b2   OR   b1,b2 /   (arrows point downwards)
+//     \ | /               |            |  /
+//      c1                c1           c1
+*/
 func TestPgraphGrouping18(t *testing.T) {
 	g1, _ := pgraph.NewGraph("g1") // original graph
 	{
@@ -801,23 +843,38 @@ func TestPgraphGrouping18(t *testing.T) {
 		g1.AddEdge(a2, c1, e3)
 		g1.AddEdge(b2, c1, e4)
 	}
-	g2, _ := pgraph.NewGraph("g2") // expected result
+	//g2, _ := pgraph.NewGraph("g2") // expected result
+	//{
+	//	a := NewNoopResTest("a1,a2")
+	//	b := NewNoopResTest("b1,b2")
+	//	c1 := NewNoopResTest("c1")
+	//	e1 := NE("e1,e3")
+	//	e2 := NE("e2,e3,e4") // e3 gets "merged through" to BOTH edges!
+	//	g2.AddEdge(a, b, e1)
+	//	g2.AddEdge(b, c1, e2)
+	//}
+	//runGraphCmp(t, g1, g2)
+	g3, _ := pgraph.NewGraph("g3") // alternative expected result
 	{
 		a := NewNoopResTest("a1,a2")
 		b := NewNoopResTest("b1,b2")
 		c1 := NewNoopResTest("c1")
-		e1 := NE("e1,e3")
-		e2 := NE("e2,e3,e4") // e3 gets "merged through" to BOTH edges!
-		g2.AddEdge(a, b, e1)
-		g2.AddEdge(b, c1, e2)
+		e1 := NE("e1")
+		e2 := NE("e2,e4")
+		e3 := NE("e3")
+		g3.AddEdge(a, b, e1)
+		g3.AddEdge(b, c1, e2)
+		g3.AddEdge(a, c1, e3)
 	}
-	runGraphCmp(t, g1, g2)
+	runGraphCmp(t, g1, g3)
 }
 
+/*
 // connected merge 0, (no change!)
 // a1            a1
 //   \     >>>     \     (arrows point downwards)
 //    a2            a2
+*/
 func TestPgraphGroupingConnected0(t *testing.T) {
 	g1, _ := pgraph.NewGraph("g1") // original graph
 	{
@@ -836,12 +893,14 @@ func TestPgraphGroupingConnected0(t *testing.T) {
 	runGraphCmp(t, g1, g2)
 }
 
+/*
 // connected merge 1, (no change!)
 // a1              a1
 //   \               \
 //    b      >>>      b      (arrows point downwards)
 //     \               \
 //      a2              a2
+*/
 func TestPgraphGroupingConnected1(t *testing.T) {
 	g1, _ := pgraph.NewGraph("g1") // original graph
 	{

engine/graph/autogroup/base.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,7 +13,19 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package autogroup
 
@@ -43,8 +55,24 @@ func (ag *baseGrouper) Init(g *pgraph.Graph) error {
 	if ag.graph != nil {
 		return fmt.Errorf("the init method has already been called")
 	}
-	ag.graph = g                            // pointer
-	ag.vertices = ag.graph.VerticesSorted() // cache in deterministic order!
+	ag.graph = g // pointer
+
+	// We sort deterministically, first by kind, and then by name. In
+	// particular, longer kind chunks sort first. So http:ui:text should
+	// appear before http:server and http:ui. This is a hack so that if we
+	// are doing hierarchical automatic grouping, it gives the http:ui:text
+	// a chance to get grouped into http:ui, before http:ui gets grouped
+	// into http:server, because once that happens, http:ui:text will never
+	// get grouped, and this won't work properly. This works, because when
+	// we start comparing iteratively the list of resources, it does this
+	// with a O(n^2) loop that compares the X and Y zero indexes first, and
+	// and then continues along. If the "longer" resources appear first,
+	// then they'll group together first. We should probably put this into
+	// a new Grouper struct, but for now we might as well leave it here.
+	//vertices := ag.graph.VerticesSorted() // formerly
+	vertices := RHVSort(ag.graph.Vertices())
+
+	ag.vertices = vertices // cache in deterministic order!
 	ag.i = 0
 	ag.j = 0
 	if len(ag.vertices) == 0 { // empty graph
@@ -89,6 +117,19 @@ func (ag *baseGrouper) VertexNext() (v1, v2 pgraph.Vertex, err error) {
 			ag.done = true
 		}
 	}
+	// TODO: is this index swap better or even valid?
+	//if ag.i < l {
+	//	ag.i++
+	//}
+	//if ag.i == l {
+	//	ag.i = 0
+	//	if ag.j < l {
+	//		ag.j++
+	//	}
+	//	if ag.j == l {
+	//		ag.done = true
+	//	}
+	//}
 
 	return
 }
@@ -110,7 +151,7 @@ func (ag *baseGrouper) VertexMerge(v1, v2 pgraph.Vertex) (v pgraph.Vertex, err e
 	return nil, fmt.Errorf("vertexMerge needs to be overridden")
 }
 
-// EdgeMerge can be overridden, since it just simple returns the first edge.
+// EdgeMerge can be overridden, since it just simply returns the first edge.
 func (ag *baseGrouper) EdgeMerge(e1, e2 pgraph.Edge) pgraph.Edge {
 	return e1 // noop
 }

engine/graph/autogroup/nonreachability.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,14 +13,25 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package autogroup
 
 import (
 	"github.com/purpleidea/mgmt/pgraph"
-
-	errwrap "github.com/pkg/errors"
+	"github.com/purpleidea/mgmt/util/errwrap"
 )
 
 // NonReachabilityGrouper is the most straight-forward algorithm for grouping.

engine/graph/autogroup/util.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,19 +13,35 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package autogroup
 
 import (
-	"github.com/purpleidea/mgmt/pgraph"
+	"fmt"
+	"sort"
+	"strings"
 
-	errwrap "github.com/pkg/errors"
+	"github.com/purpleidea/mgmt/engine"
+	"github.com/purpleidea/mgmt/pgraph"
+	"github.com/purpleidea/mgmt/util/errwrap"
 )
 
-// VertexMerge merges v2 into v1 by reattaching the edges where appropriate,
-// and then by deleting v2 from the graph. Since more than one edge between two
-// vertices is not allowed, duplicate edges are merged as well. an edge merge
+// VertexMerge merges v2 into v1 by reattaching the edges where appropriate, and
+// then by deleting v2 from the graph. Since more than one edge between two
+// vertices is not allowed, duplicate edges are merged as well. An edge merge
 // function can be provided if you'd like to control how you merge the edges!
 func VertexMerge(g *pgraph.Graph, v1, v2 pgraph.Vertex, vertexMergeFn func(pgraph.Vertex, pgraph.Vertex) (pgraph.Vertex, error), edgeMergeFn func(pgraph.Edge, pgraph.Edge) pgraph.Edge) error {
 	// methodology
@@ -113,8 +129,17 @@ func VertexMerge(g *pgraph.Graph, v1, v2 pgraph.Vertex, vertexMergeFn func(pgrap
 			// note: This branch isn't used if the vertexMergeFn
 			// decides to just merge logically on its own instead
 			// of actually returning something that we then merge.
-			v1 = v // TODO: ineffassign?
+			v1 = v // XXX: ineffassign?
 			//*v1 = *v
+
+			// Ensure that everything still validates. (For safety!)
+			r, ok := v1.(engine.Res) // TODO: v ?
+			if !ok {
+				return fmt.Errorf("not a Res")
+			}
+			if err := engine.Validate(r); err != nil {
+				return errwrap.Wrapf(err, "the Res did not Validate")
+			}
 		}
 	}
 	g.DeleteVertex(v2) // remove grouped vertex
@@ -125,3 +150,67 @@ func VertexMerge(g *pgraph.Graph, v1, v2 pgraph.Vertex, vertexMergeFn func(pgrap
 	}
 	return nil // success
 }
+
+// RHVSlice is a linear list of vertices. It can be sorted by the Kind, taking
+// into account the hierarchy of names separated by colons. Afterwards, it uses
+// String() to avoid the non-determinism in the map type. RHV stands for Reverse
+// Hierarchical Vertex, meaning the hierarchical topology of the vertex
+// (resource) names are used.
+type RHVSlice []pgraph.Vertex
+
+// Len returns the length of the slice of vertices.
+func (obj RHVSlice) Len() int { return len(obj) }
+
+// Swap swaps two elements in the slice.
+func (obj RHVSlice) Swap(i, j int) { obj[i], obj[j] = obj[j], obj[i] }
+
+// Less returns the smaller element in the sort order according to the
+// aforementioned rules.
+// XXX: Add some tests to make sure I didn't get any "reverse" part backwards.
+func (obj RHVSlice) Less(i, j int) bool {
+	resi, oki := obj[i].(engine.Res)
+	resj, okj := obj[j].(engine.Res)
+	if !oki || !okj || resi.Kind() == "" || resj.Kind() == "" {
+		// One of these isn't a normal Res, so just compare normally.
+		return obj[i].String() > obj[j].String() // reverse
+	}
+
+	si := strings.Split(resi.Kind(), ":")
+	sj := strings.Split(resj.Kind(), ":")
+	// both lengths should each be at least one now
+	li := len(si)
+	lj := len(sj)
+
+	if li != lj { // eg: http:ui vs. http:ui:text
+		return li > lj // reverse
+	}
+
+	// same number of chunks
+	for k := 0; k < li; k++ {
+		if si[k] != sj[k] { // lhs chunk differs
+			return si[k] > sj[k] // reverse
+		}
+
+		// if the chunks are the same, we continue...
+	}
+
+	// They must all have the same chunks, so finally we compare the names.
+	return resi.Name() > resj.Name() // reverse
+}
+
+// Sort is a convenience method.
+func (obj RHVSlice) Sort() { sort.Sort(obj) }
+
+// RHVSort returns a deterministically sorted slice of all vertices in the list.
+// The order is sorted by the Kind, taking into account the hierarchy of names
+// separated by colons. Afterwards, it uses String() to avoid the
+// non-determinism in the map type. RHV stands for Reverse Hierarchical Vertex,
+// meaning the hierarchical topology of the vertex (resource) names are used.
+func RHVSort(vertices []pgraph.Vertex) []pgraph.Vertex {
+	var vs []pgraph.Vertex
+	for _, v := range vertices { // copy first
+		vs = append(vs, v)
+	}
+	sort.Sort(RHVSlice(vs)) // add determinism
+	return vs
+}

engine/graph/engine.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,8 +13,23 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
+// Package graph contains the actual implementation of the resource graph engine
+// that runs the graph of resources in real-time. This package has the algorithm
+// that runs all the graph transitions.
 package graph
 
 import (
@@ -25,38 +40,50 @@ import (
 
 	"github.com/purpleidea/mgmt/converger"
 	"github.com/purpleidea/mgmt/engine"
-	"github.com/purpleidea/mgmt/engine/event"
+	"github.com/purpleidea/mgmt/engine/local"
+	engineUtil "github.com/purpleidea/mgmt/engine/util"
 	"github.com/purpleidea/mgmt/pgraph"
+	"github.com/purpleidea/mgmt/util/errwrap"
 	"github.com/purpleidea/mgmt/util/semaphore"
+)
 
-	multierr "github.com/hashicorp/go-multierror"
-	errwrap "github.com/pkg/errors"
+const (
+	// StateDir is the name of the sub directory where all the local
+	// resource state is stored.
+	StateDir = "state"
 )
 
 // Engine encapsulates a generic graph and manages its operations.
 type Engine struct {
 	Program  string
+	Version  string
 	Hostname string
-	World    engine.World
+
+	Converger *converger.Coordinator
+	Local     *local.API
+	World     engine.World
 
 	// Prefix is a unique directory prefix which can be used. It should be
 	// created if needed.
-	Prefix    string
-	Converger converger.Converger
-
-	Debug bool
-	Logf  func(format string, v ...interface{})
+	Prefix string
+	Debug  bool
+	Logf   func(format string, v ...interface{})
 
 	graph     *pgraph.Graph
 	nextGraph *pgraph.Graph
 	state     map[pgraph.Vertex]*State
-	waits     map[pgraph.Vertex]*sync.WaitGroup
+	waits     map[pgraph.Vertex]*sync.WaitGroup // wg for the Worker func
+	wlock     *sync.Mutex                       // lock around waits map
+
+	mlock *sync.Mutex                            // metas lock
+	metas map[engine.ResPtrUID]*engine.MetaState // meta state
 
 	slock *sync.Mutex // semaphore lock
 	semas map[string]*semaphore.Semaphore
 
-	wg *sync.WaitGroup
+	wg *sync.WaitGroup // wg for the whole engine (only used for close)
 
+	paused    bool // are we paused?
 	fastPause bool
 }
 
@@ -64,6 +91,13 @@ type Engine struct {
 // If the struct does not validate, or it cannot initialize, then this errors.
 // Initially it will contain an empty graph.
 func (obj *Engine) Init() error {
+	if obj.Program == "" {
+		return fmt.Errorf("the Program is empty")
+	}
+	if obj.Hostname == "" {
+		return fmt.Errorf("the Hostname is empty")
+	}
+
 	var err error
 	if obj.graph, err = pgraph.NewGraph("graph"); err != nil {
 		return err
@@ -78,12 +112,18 @@ func (obj *Engine) Init() error {
 
 	obj.state = make(map[pgraph.Vertex]*State)
 	obj.waits = make(map[pgraph.Vertex]*sync.WaitGroup)
+	obj.wlock = &sync.Mutex{}
+
+	obj.mlock = &sync.Mutex{}
+	obj.metas = make(map[engine.ResPtrUID]*engine.MetaState)
 
 	obj.slock = &sync.Mutex{}
 	obj.semas = make(map[string]*semaphore.Semaphore)
 
 	obj.wg = &sync.WaitGroup{}
 
+	obj.paused = true // start off true, so we can Resume after first Commit
+
 	return nil
 }
 
@@ -118,14 +158,14 @@ func (obj *Engine) Validate() error {
 		}
 
 		if err := engine.Validate(res); err != nil {
-			return errwrap.Wrapf(err, "the Res did not Validate")
+			return fmt.Errorf("%s did not Validate: %v", res, err)
 		}
 	}
 	return nil
 }
 
 // Apply a function to the pending graph. You must pass in a function which will
-// receive this graph as input, and return an error if it something does not
+// receive this graph as input, and return an error if something does not
 // succeed.
 func (obj *Engine) Apply(fn func(*pgraph.Graph) error) error {
 	return fn(obj.nextGraph)
@@ -135,8 +175,22 @@ func (obj *Engine) Apply(fn func(*pgraph.Graph) error) error {
 // it errors, then the running graph wasn't changed. It is recommended that you
 // pause the engine before running this, and resume it after you're done.
 func (obj *Engine) Commit() error {
+	// It would be safer to lock this, but it would be slower and mask bugs.
+	//obj.mutex.Lock()
+	//defer obj.mutex.Unlock()
+
 	// TODO: Does this hurt performance or graph changes ?
 
+	activeMetas := make(map[engine.ResPtrUID]struct{})
+	for vertex := range obj.state {
+		res, ok := vertex.(engine.Res)
+		if !ok { // should not happen, previously validated
+			return fmt.Errorf("not a Res")
+		}
+		activeMetas[engine.PtrUID(res)] = struct{}{} // add
+	}
+
+	start := []func() error{} // functions to run after graphsync to start...
 	vertexAddFn := func(vertex pgraph.Vertex) error {
 		// some of these validation steps happen before this Commit step
 		// in Validate() to avoid erroring here. These are redundant.
@@ -153,20 +207,22 @@ func (obj *Engine) Commit() error {
 			return fmt.Errorf("the Res state already exists")
 		}
 
+		activeMetas[engine.PtrUID(res)] = struct{}{} // add
+
 		if obj.Debug {
 			obj.Logf("Validate(%s)", res)
 		}
 		err := engine.Validate(res)
 		if obj.Debug {
-			obj.Logf("Validate(%s): Return(%+v)", res, err)
+			obj.Logf("Validate(%s): Return(%s)", res, engineUtil.CleanError(err))
 		}
 		if err != nil {
 			return errwrap.Wrapf(err, "the Res did not Validate")
 		}
 
-		// FIXME: is res.Name() sufficiently unique to use as a UID here?
-		pathUID := fmt.Sprintf("%s-%s", res.Kind(), res.Name())
-		statePrefix := fmt.Sprintf("%s/", path.Join(obj.Prefix, "state", pathUID))
+		pathUID := engineUtil.ResPathUID(res)
+		statePrefix := fmt.Sprintf("%s/", path.Join(obj.statePrefix(), pathUID))
+
 		// don't create this unless it *will* be used
 		//if err := os.MkdirAll(statePrefix, 0770); err != nil {
 		//	return errwrap.Wrapf(err, "can't create state prefix")
@@ -174,15 +230,17 @@ func (obj *Engine) Commit() error {
 
 		obj.waits[vertex] = &sync.WaitGroup{}
 		obj.state[vertex] = &State{
-			//Graph: obj.graph, // TODO: what happens if we swap the graph?
+			Graph:  obj.graph, // Update if we swap the graph!
 			Vertex: vertex,
 
 			Program:  obj.Program,
+			Version:  obj.Version,
 			Hostname: obj.Hostname,
 
+			//Converger: obj.Converger,
+			Local:  obj.Local,
 			World:  obj.World,
 			Prefix: statePrefix,
-			//Converger: obj.Converger,
 
 			Debug: obj.Debug,
 			Logf: func(format string, v ...interface{}) {
@@ -192,94 +250,210 @@ func (obj *Engine) Commit() error {
 		if err := obj.state[vertex].Init(); err != nil {
 			return errwrap.Wrapf(err, "the Res did not Init")
 		}
+
+		fn := func() error {
+			// start the Worker
+			obj.wg.Add(1)
+			obj.wlock.Lock()
+			obj.waits[vertex].Add(1)
+			obj.wlock.Unlock()
+			go func(v pgraph.Vertex) {
+				defer obj.wg.Done()
+				defer func() {
+					// we need this lock, because this go
+					// routine could run when the next fn
+					// function above here is running...
+					obj.wlock.Lock()
+					obj.waits[v].Done()
+					obj.wlock.Unlock()
+				}()
+
+				if obj.Debug || true {
+					obj.Logf("%s: Working...", v)
+				}
+				// contains the Watch and CheckApply loops
+				err := obj.Worker(v)
+				if obj.Debug || true {
+					if s := engineUtil.CleanError(err); err != nil {
+						obj.Logf("%s: Error: %s", v, s)
+					} else {
+						obj.Logf("%s: Exited...", v)
+					}
+				}
+				obj.state[v].workerErr = err // store the error
+				// If the Rewatch metaparam is true, then this will get
+				// restarted if we do a graph cmp swap. This is why the
+				// graph cmp function runs the removes before the adds.
+				// XXX: This should feed into an $error var in the lang.
+			}(vertex)
+			return nil
+		}
+		start = append(start, fn) // do this at the end, if it's needed
 		return nil
 	}
+
+	free := []func() error{} // functions to run after graphsync to reset...
 	vertexRemoveFn := func(vertex pgraph.Vertex) error {
+		res, ok := vertex.(engine.Res)
+		if !ok { // should not happen, previously validated
+			return fmt.Errorf("not a Res")
+		}
+		delete(activeMetas, engine.PtrUID(res))
+
 		// wait for exit before starting new graph!
-		obj.state[vertex].Event(event.EventExit) // signal an exit
-		obj.waits[vertex].Wait()                 // sync
+		close(obj.state[vertex].removeDone)   // causes doneCtx to cancel
+		close(obj.state[vertex].resumeSignal) // unblock (it only closes here)
+		obj.waits[vertex].Wait()              // sync
 
 		// close the state and resource
 		// FIXME: will this mess up the sync and block the engine?
-		if err := obj.state[vertex].Close(); err != nil {
-			return errwrap.Wrapf(err, "the Res did not Close")
+		if err := obj.state[vertex].Cleanup(); err != nil {
+			return errwrap.Wrapf(err, "the Res did not Cleanup")
 		}
 
 		// delete to free up memory from old graphs
-		delete(obj.state, vertex)
-		delete(obj.waits, vertex)
+		fn := func() error {
+			delete(obj.state, vertex)
+			delete(obj.waits, vertex)
+			return nil
+		}
+		free = append(free, fn) // do this at the end, so we don't panic
 		return nil
 	}
 
+	// add the Worker swap (reload) on error decision into this vertexCmpFn
+	vertexCmpFn := func(v1, v2 pgraph.Vertex) (bool, error) {
+		r1, ok1 := v1.(engine.Res)
+		r2, ok2 := v2.(engine.Res)
+		if !ok1 || !ok2 { // should not happen, previously validated
+			return false, fmt.Errorf("not a Res")
+		}
+		m1 := r1.MetaParams()
+		m2 := r2.MetaParams()
+		swap1, swap2 := true, true // assume default of true
+		if m1 != nil {
+			swap1 = m1.Rewatch
+		}
+		if m2 != nil {
+			swap2 = m2.Rewatch
+		}
+
+		s1, ok1 := obj.state[v1]
+		s2, ok2 := obj.state[v2]
+		x1, x2 := false, false
+		if ok1 {
+			x1 = s1.workerErr != nil && swap1
+		}
+		if ok2 {
+			x2 = s2.workerErr != nil && swap2
+		}
+
+		if x1 || x2 {
+			// We swap, even if they're the same, so that we reload!
+			// This causes an add and remove of the "same" vertex...
+			return false, nil
+		}
+
+		return engine.VertexCmpFn(v1, v2) // do the normal cmp otherwise
+	}
+
 	// If GraphSync succeeds, it updates the receiver graph accordingly...
 	// Running the shutdown in vertexRemoveFn does not need to happen in a
 	// topologically sorted order because it already paused in that order.
 	obj.Logf("graph sync...")
-	if err := obj.graph.GraphSync(obj.nextGraph, engine.VertexCmpFn, vertexAddFn, vertexRemoveFn, engine.EdgeCmpFn); err != nil {
+	if err := obj.graph.GraphSync(obj.nextGraph, vertexCmpFn, vertexAddFn, vertexRemoveFn, engine.EdgeCmpFn); err != nil {
 		return errwrap.Wrapf(err, "error running graph sync")
 	}
+
+	// This happens after GraphSync when vertexRemoveFn and vertexAddFn are
+	// done running. Those two modified the activeMetas map. It's important
+	// that vertexRemoveFn runs before vertexAddFn, but GraphSync guarantees
+	// that, and it would be kind of illogical to not run things that way.
+	metaGC := make(map[engine.ResPtrUID]struct{}) // which metas should we garbage collect?
+	obj.mlock.Lock()
+	for ptrUID := range obj.metas {
+		if _, exists := activeMetas[ptrUID]; !exists {
+			metaGC[ptrUID] = struct{}{}
+		}
+	}
+	for ptrUID := range metaGC {
+		delete(obj.metas, ptrUID) // otherwise, this could grow forever
+	}
+	obj.mlock.Unlock()
+
+	// We run these afterwards, so that we don't unnecessarily start anyone
+	// if GraphSync failed in some way. Otherwise we'd have to do clean up!
+	for _, fn := range start {
+		if err := fn(); err != nil {
+			return errwrap.Wrapf(err, "error running start fn")
+		}
+	}
+	// We run these afterwards, so that the state structs (that might get
+	// referenced) are not destroyed while someone might poke or use one.
+	for _, fn := range free {
+		if err := fn(); err != nil {
+			return errwrap.Wrapf(err, "error running free fn")
+		}
+	}
 	obj.nextGraph = nil
 
 	// After this point, we must not error or we'd need to restore all of
 	// the changes that we'd made to the previously primary graph. This is
 	// because this function is meant to atomically swap the graphs safely.
 
-	// TODO: update all the `State` structs with the new Graph pointer
-	//for _, vertex := range obj.graph.Vertices() {
-	//	state, exists := obj.state[vertex]
-	//	if !exists {
-	//		continue
-	//	}
-	//	state.Graph = obj.graph // update pointer to graph
-	//}
+	// Update all the `State` structs with the new Graph pointer.
+	for _, vertex := range obj.graph.Vertices() {
+		state, exists := obj.state[vertex]
+		if !exists {
+			continue
+		}
+		state.Graph = obj.graph // update pointer to graph
+	}
 
 	return nil
 }
 
-// Start runs the currently active graph. It also un-pauses the graph if it was
-// paused.
-func (obj *Engine) Start() error {
+// Resume runs the currently active graph. It also un-pauses the graph if it was
+// paused. Very little that is interesting should happen here. It all happens in
+// the Commit method. After Commit, new things are already started, but we still
+// need to Resume any pre-existing resources. Do not call this concurrently with
+// the Pause method.
+func (obj *Engine) Resume() error {
+	// It would be safer to lock this, but it would be slower and mask bugs.
+	//obj.mutex.Lock()
+	//defer obj.mutex.Unlock()
+
+	if !obj.paused {
+		return fmt.Errorf("already resumed")
+	}
+
 	topoSort, err := obj.graph.TopologicalSort()
 	if err != nil {
 		return err
 	}
-	indegree := obj.graph.InDegree() // compute all of the indegree's
+	//indegree := obj.graph.InDegree() // compute all of the indegree's
 	reversed := pgraph.Reverse(topoSort)
 
 	for _, vertex := range reversed {
-		state := obj.state[vertex]
-		state.starter = (indegree[vertex] == 0)
-		var unpause = true // assume true
+		// The very first resume is skipped as those resources are
+		// already running! We could do that by checking here, but it is
+		// more convenient to just have a state struct field (paused) to
+		// track things for this instead. As a bonus, it helps us know
+		// if a resource is paused or not if we print for debugging.
+		//if !obj.state[vertex].initialStartupDone {
+		//	obj.state[vertex].initialStartupDone = true
+		//	continue
+		//}
 
-		if !state.working { // if not running...
-			state.working = true
-			unpause = false // doesn't need unpausing if starting
-			obj.wg.Add(1)
-			obj.waits[vertex].Add(1)
-			go func(v pgraph.Vertex) {
-				defer obj.wg.Done()
-				defer obj.waits[vertex].Done()
-				defer func() {
-					obj.state[v].working = false
-				}()
-
-				obj.Logf("Worker(%s)", v)
-				// contains the Watch and CheckApply loops
-				err := obj.Worker(v)
-				obj.Logf("Worker(%s): Exited(%+v)", v, err)
-			}(vertex)
-		}
-
-		select {
-		case <-state.started:
-		case <-state.stopped: // we failed on Watch start
-		}
-
-		if unpause { // unpause (if needed)
-			obj.state[vertex].Event(event.EventStart)
-		}
+		//obj.state[vertex].starter = (indegree[vertex] == 0)
+		obj.state[vertex].Resume() // doesn't error
+		// This always works because if a resource errored while it was
+		// paused, then we're in the paused state and we can still exit
+		// from there. If a resource errors when we're trying to Pause
+		// then it will only succeed without error if the resource ACKs.
 	}
 	// we wait for everyone to start before exiting!
+	obj.paused = false
 	return nil
 }
 
@@ -290,40 +464,52 @@ func (obj *Engine) Start() error {
 // This is because once you've started a fast pause, some dependencies might
 // have been skipped when fast pausing, and future resources might have missed a
 // poke. In general this is only called when you're trying to hurry up the exit.
+// XXX: Not implemented
 func (obj *Engine) SetFastPause() {
 	obj.fastPause = true
 }
 
-// Pause the active, running graph. At the moment this cannot error.
-func (obj *Engine) Pause(fastPause bool) {
+// Pause the active, running graph.
+func (obj *Engine) Pause(fastPause bool) error {
+	// It would be safer to lock this, but it would be slower and mask bugs.
+	//obj.mutex.Lock()
+	//defer obj.mutex.Unlock()
+
+	if obj.paused {
+		return fmt.Errorf("already paused")
+	}
+
 	obj.fastPause = fastPause
 	topoSort, _ := obj.graph.TopologicalSort()
 	for _, vertex := range topoSort { // squeeze out the events...
 		// The Event is sent to an unbuffered channel, so this event is
 		// synchronous, and as a result it blocks until it is received.
-		obj.state[vertex].Event(event.EventPause)
+		if err := obj.state[vertex].Pause(); err != nil && err != engine.ErrClosed {
+			return err
+		}
 	}
 
+	obj.paused = true
+
 	// we are now completely paused...
 	obj.fastPause = false // reset
+	return nil
 }
 
-// Close triggers a shutdown. Engine must be already paused before this is run.
-func (obj *Engine) Close() error {
-	var reterr error
-
-	emptyGraph, err := pgraph.NewGraph("empty")
-	if err != nil {
-		reterr = multierr.Append(reterr, err) // list of errors
-	}
+// Shutdown the engine. Engine must be already paused before this is run. It is
+// actually just a Load of an empty graph and a Commit. It waits for all the
+// resources to exit before returning.
+func (obj *Engine) Shutdown() error {
+	emptyGraph, reterr := pgraph.NewGraph("empty")
 
 	// this is a graph switch (graph sync) that switches to an empty graph!
 	if err := obj.Load(emptyGraph); err != nil { // copy in empty graph
-		reterr = multierr.Append(reterr, err)
+		reterr = errwrap.Append(reterr, err)
 	}
+	// FIXME: Do we want to run commit if Load failed? Does this even work?
 	// the commit will cause the graph sync to shut things down cleverly...
 	if err := obj.Commit(); err != nil {
-		reterr = multierr.Append(reterr, err)
+		reterr = errwrap.Append(reterr, err)
 	}
 
 	obj.wg.Wait() // for now, this doesn't need to be a separate Wait() method
@@ -334,3 +520,8 @@ func (obj *Engine) Close() error {
 func (obj *Engine) Graph() *pgraph.Graph {
 	return obj.graph
 }
+
+// statePrefix returns the dir where all the resource state is stored locally.
+func (obj *Engine) statePrefix() string {
+	return fmt.Sprintf("%s/", path.Join(obj.Prefix, StateDir))
+}

engine/graph/graph_test.go

@@ -0,0 +1,49 @@
+// Mgmt
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
+
+//go:build !root
+
+package graph
+
+import (
+	"fmt"
+	"testing"
+
+	"github.com/purpleidea/mgmt/util/errwrap"
+)
+
+func TestMultiErr(t *testing.T) {
+	var err error
+	e := fmt.Errorf("some error")
+	err = errwrap.Append(err, e) // build an error from a nil base
+	// ensure that this lib allows us to append to a nil
+	if err == nil {
+		t.Errorf("missing error")
+	}
+}

engine/graph/refresh.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,7 +13,19 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package graph
 

engine/graph/reverse.go

@@ -0,0 +1,311 @@
+// Mgmt
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
+
+package graph
+
+import (
+	"fmt"
+	"os"
+	"path"
+	"sort"
+
+	"github.com/purpleidea/mgmt/engine"
+	engineUtil "github.com/purpleidea/mgmt/engine/util"
+	"github.com/purpleidea/mgmt/pgraph"
+	"github.com/purpleidea/mgmt/util/errwrap"
+)
+
+const (
+	// ReverseFile is the file name in the resource state dir where any
+	// reversal information is stored.
+	ReverseFile = "reverse"
+
+	// ReversePerm is the permissions mode used to create the ReverseFile.
+	ReversePerm = 0600
+)
+
+// Reversals adds the reversals onto the loaded graph. This should happen last,
+// and before Commit.
+func (obj *Engine) Reversals() error {
+	if obj.nextGraph == nil {
+		return fmt.Errorf("there is no active graph to add reversals to")
+	}
+
+	// Initially get all of the reversals to seek out all possible errors.
+	// XXX: The engine needs to know where data might have been stored if we
+	// XXX: want to potentially allow alternate read/write paths, like etcd.
+	// XXX: In this scenario, we'd have to store a token somewhere to let us
+	// XXX: know to look elsewhere for the special ReversalList read method.
+	data, err := obj.ReversalList() // (map[string]string, error)
+	if err != nil {
+		return errwrap.Wrapf(err, "the reversals had errors")
+	}
+
+	if len(data) == 0 {
+		return nil // end early
+	}
+
+	resMatch := func(r1, r2 engine.Res) bool { // simple match on UID only!
+		if r1.Kind() != r2.Kind() {
+			return false
+		}
+		if r1.Name() != r2.Name() {
+			return false
+		}
+		return true
+	}
+	resInList := func(needle engine.Res, haystack []engine.Res) bool {
+		for _, res := range haystack {
+			if resMatch(needle, res) {
+				return true
+			}
+		}
+		return false
+	}
+
+	if obj.Debug {
+		obj.Logf("decoding %d reversals...", len(data))
+	}
+	resources := []engine.Res{}
+
+	// do this in a sorted order so that it errors deterministically
+	sorted := []string{}
+	for key := range data {
+		sorted = append(sorted, key)
+	}
+	sort.Strings(sorted)
+	for _, key := range sorted {
+		val := data[key]
+		// XXX: replace this ResToB64 method with one that stores it in
+		// a human readable format, in case someone wants to hack and
+		// edit it manually.
+		// XXX: we probably want this to be YAML, it works with the diff
+		// too...
+		r, err := engineUtil.B64ToRes(val)
+		if err != nil {
+			return errwrap.Wrapf(err, "error decoding res with UID: `%s`", key)
+		}
+
+		res, ok := r.(engine.ReversibleRes)
+		if !ok {
+			// this requirement is here to keep things simpler...
+			return errwrap.Wrapf(err, "decoded res with UID: `%s` was not reversible", key)
+		}
+
+		matchFn := func(vertex pgraph.Vertex) (bool, error) {
+			r, ok := vertex.(engine.Res)
+			if !ok {
+				return false, fmt.Errorf("not a Res")
+			}
+			if !resMatch(r, res) {
+				return false, nil
+			}
+			return true, nil
+		}
+
+		// FIXME: not efficient, we could build a cache-map first
+		vertex, err := obj.nextGraph.VertexMatchFn(matchFn) // (Vertex, error)
+		if err != nil {
+			return errwrap.Wrapf(err, "error searching graph for match")
+		}
+		if vertex != nil { // found one!
+			continue // it doesn't need reversing yet
+		}
+
+		// TODO: check for (incompatible?) duplicates instead
+		if resInList(res, resources) { // we've already got this one...
+			continue
+		}
+
+		// We set this in two different places to be safe. It ensures
+		// that we erase the reversal state file after we've used it.
+		res.ReversibleMeta().Reversal = true // set this for later...
+
+		resources = append(resources, res)
+	}
+
+	if len(resources) == 0 {
+		return nil // end early
+	}
+
+	// Now that we've passed the chance of any errors, we modify the graph.
+	obj.Logf("adding %d reversals...", len(resources))
+	for _, res := range resources {
+		obj.nextGraph.AddVertex(res)
+	}
+	// TODO: Do we want a way for stored reversals to add edges too?
+
+	// It would be great to ensure we didn't add any graph cycles here, but
+	// instead of checking now, we'll move the check into the main loop.
+	return nil
+}
+
+// ReversalList returns all the available pending reversal data on this host. It
+// can then be decoded by whatever method is appropriate for.
+func (obj *Engine) ReversalList() (map[string]string, error) {
+	result := make(map[string]string) // some key to contents
+
+	dir := obj.statePrefix() // loop through this dir...
+	files, err := os.ReadDir(dir)
+	if err != nil && !os.IsNotExist(err) {
+		return nil, errwrap.Wrapf(err, "error reading list of state dirs")
+	} else if err != nil {
+		return result, nil // nothing found, no state dir exists yet
+	}
+
+	for _, x := range files {
+		key := x.Name() // some uid for the resource
+		file := path.Join(dir, key, ReverseFile)
+		content, err := os.ReadFile(file)
+		if err != nil && !os.IsNotExist(err) {
+			return nil, errwrap.Wrapf(err, "could not read reverse file: %s", file)
+		} else if err != nil {
+			continue // file does not exist, skip
+		}
+
+		// file exists!
+		str := string(content)
+		result[key] = str // save
+	}
+
+	return result, nil
+}
+
+// ReversalInit performs the reversal initialization steps if necessary for this
+// resource.
+func (obj *State) ReversalInit() error {
+	res, ok := obj.Vertex.(engine.ReversibleRes)
+	if !ok {
+		return nil // nothing to do
+	}
+
+	if res.ReversibleMeta().Disabled {
+		return nil // nothing to do, reversal isn't enabled
+	}
+
+	// If the reversal is enabled, but we are the result of a previous
+	// reversal, then this will overwrite that older reversal request, and
+	// our resource should be designed to deal with that. This happens if we
+	// return a reversible resource as the reverse of a resource that was
+	// reversed. It's probably fairly rare.
+	if res.ReversibleMeta().Reversal {
+		obj.Logf("triangle reversal") // warn!
+	}
+
+	r, err := res.Reversed()
+	if err != nil {
+		return errwrap.Wrapf(err, "could not reverse: %s", res.String())
+	}
+	if r == nil {
+		return nil // this can't be reversed, or isn't implemented here
+	}
+
+	// We set this in two different places to be safe. It ensures that we
+	// erase the reversal state file after we've used it.
+	r.ReversibleMeta().Reversal = true // set this for later...
+
+	// XXX: replace this ResToB64 method with one that stores it in a human
+	// readable format, in case someone wants to hack and edit it manually.
+	// XXX: we probably want this to be YAML, it works with the diff too...
+	str, err := engineUtil.ResToB64(r)
+	if err != nil {
+		return errwrap.Wrapf(err, "could not encode: %s", res.String())
+	}
+
+	// TODO: put this method on traits.Reversible as part of the interface?
+	return obj.ReversalWrite(str, res.ReversibleMeta().Overwrite) // Store!
+}
+
+// ReversalCleanup performs the reversal shutdown steps if necessary for this
+// resource.
+func (obj *State) ReversalCleanup() error {
+	res, ok := obj.Vertex.(engine.ReversibleRes)
+	if !ok {
+		return nil // nothing to do
+	}
+
+	// Don't check res.ReversibleMeta().Disabled because we're removing the
+	// previous one. That value only applies if we're doing a new reversal.
+
+	if !res.ReversibleMeta().Reversal {
+		return nil // nothing to erase, we're not a reversal resource
+	}
+
+	if !obj.isStateOK.Load() { // did we successfully reverse? (mutex RLock/RUnlock)
+		obj.Logf("did not complete reversal") // warn
+		return nil
+	}
+
+	// TODO: put this method on traits.Reversible as part of the interface?
+	return obj.ReversalDelete() // Erase our reversal instructions.
+}
+
+// ReversalWrite stores the reversal state information for this resource.
+func (obj *State) ReversalWrite(str string, overwrite bool) error {
+	dir, err := obj.varDir("") // private version
+	if err != nil {
+		return errwrap.Wrapf(err, "could not get VarDir for reverse")
+	}
+	file := path.Join(dir, ReverseFile) // return a unique file
+
+	content, err := os.ReadFile(file)
+	if err != nil && !os.IsNotExist(err) {
+		return errwrap.Wrapf(err, "could not read reverse file: %s", file)
+	}
+
+	// file exists and we shouldn't overwrite if different
+	if err == nil && !overwrite {
+		// compare to existing file
+		oldStr := string(content)
+		if str != oldStr {
+			obj.Logf("existing, pending, reversible resource exists")
+			//obj.Logf("diff:")
+			//obj.Logf("") // TODO: print the diff w/o secret values
+			return fmt.Errorf("existing, pending, reversible resource exists")
+		}
+	}
+
+	return os.WriteFile(file, []byte(str), ReversePerm)
+}
+
+// ReversalDelete removes the reversal state information for this resource.
+func (obj *State) ReversalDelete() error {
+	dir, err := obj.varDir("") // private version
+	if err != nil {
+		return errwrap.Wrapf(err, "could not get VarDir for reverse")
+	}
+	file := path.Join(dir, ReverseFile) // return a unique file
+
+	// FIXME: why do we see these removals when there isn't a state file?
+	if err = os.Remove(file); os.IsNotExist(err) {
+		return nil // ignore missing files
+	}
+
+	return errwrap.Wrapf(err, "could not remove reverse state file")
+}

engine/graph/semaphore.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,7 +13,19 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package graph
 
@@ -23,9 +35,8 @@ import (
 	"strconv"
 	"strings"
 
+	"github.com/purpleidea/mgmt/util/errwrap"
 	"github.com/purpleidea/mgmt/util/semaphore"
-
-	multierr "github.com/hashicorp/go-multierror"
 )
 
 // SemaSep is the trailing separator to split the semaphore id from the size.
@@ -46,9 +57,8 @@ func (obj *Engine) semaLock(semas []string) error {
 		}
 		obj.slock.Unlock()
 
-		if err := sema.P(1); err != nil { // lock!
-			reterr = multierr.Append(reterr, err) // list of errors
-		}
+		err := sema.P(1)                     // lock!
+		reterr = errwrap.Append(reterr, err) // list of errors
 	}
 	return reterr
 }
@@ -65,9 +75,8 @@ func (obj *Engine) semaUnlock(semas []string) error {
 			panic(fmt.Sprintf("graph: sema: %s does not exist", id))
 		}
 
-		if err := sema.V(1); err != nil { // unlock!
-			reterr = multierr.Append(reterr, err) // list of errors
-		}
+		err := sema.V(1)                     // unlock!
+		reterr = errwrap.Append(reterr, err) // list of errors
 	}
 	return reterr
 }

engine/graph/semaphore_test.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,9 +13,21 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
-// +build !root
+//go:build !root
 
 package graph
 

engine/graph/sendrecv.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,94 +13,267 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package graph
 
 import (
 	"fmt"
 	"reflect"
+	"sort"
 
 	"github.com/purpleidea/mgmt/engine"
-
-	multierr "github.com/hashicorp/go-multierror"
-	errwrap "github.com/pkg/errors"
+	engineUtil "github.com/purpleidea/mgmt/engine/util"
+	"github.com/purpleidea/mgmt/lang/types"
+	"github.com/purpleidea/mgmt/util/errwrap"
 )
 
+// RecvFn represents a custom Recv function which can be used in place of the
+// stock, built-in one. This is needed if we want to receive from a different
+// resource data source than our own. (Only for special occasions of course!)
+type RecvFn func(engine.RecvableRes) (map[string]*engine.Send, error)
+
 // SendRecv pulls in the sent values into the receive slots. It is called by the
 // receiver and must be given as input the full resource struct to receive on.
-// It applies the loaded values to the resource.
-func (obj *Engine) SendRecv(res engine.RecvableRes) (map[string]bool, error) {
-	recv := res.Recv()
-	if obj.Debug {
-		// NOTE: this could expose private resource data like passwords
-		obj.Logf("%s: SendRecv: %+v", res, recv)
+// It applies the loaded values to the resource. It is called recursively, as it
+// recurses into any grouped resources found within the first receiver. It
+// returns a map of resource pointer, to resource field key, to changed boolean.
+func SendRecv(res engine.RecvableRes, fn RecvFn) (map[engine.RecvableRes]map[string]*engine.Send, error) {
+	updated := make(map[engine.RecvableRes]map[string]*engine.Send) // list of updated keys
+	if groupableRes, ok := res.(engine.GroupableRes); ok {
+		for _, x := range groupableRes.GetGroup() { // grouped elements
+			recvableRes, ok := x.(engine.RecvableRes)
+			if !ok {
+				continue
+			}
+			//if obj.Debug {
+			//	obj.Logf("SendRecv: %s: grouped: %s", res, x) // receiving here
+			//}
+			// We need to recurse here so that autogrouped resources
+			// inside autogrouped resources would work... In case we
+			// work correctly. We just need to make sure that things
+			// are grouped in the correct order, but that is not our
+			// problem! Recurse and merge in the changed results...
+			innerUpdated, err := SendRecv(recvableRes, fn)
+			if err != nil {
+				return nil, errwrap.Wrapf(err, "recursive SendRecv error")
+			}
+			for r, m := range innerUpdated { // res ptr, map
+				if _, exists := updated[r]; !exists {
+					updated[r] = make(map[string]*engine.Send)
+				}
+				for s, send := range m { // map[string]*engine.Send
+					b := send.Changed
+					// don't overwrite in case one exists...
+					if old, exists := updated[r][s]; exists {
+						b = b || old.Changed // unlikely i think
+					}
+					if _, exists := updated[r][s]; !exists {
+						newSend := &engine.Send{
+							Res:     send.Res,
+							Key:     send.Key,
+							Changed: b,
+						}
+						updated[r][s] = newSend
+					}
+					updated[r][s].Changed = b
+				}
+			}
+		}
 	}
-	var updated = make(map[string]bool) // list of updated keys
+
 	var err error
-	for k, v := range recv {
-		updated[k] = false // default
-		v.Changed = false  // reset to the default
+	recv := res.Recv()
+	if fn != nil {
+		recv, err = fn(res) // use a custom Recv function
+		if err != nil {
+			return nil, err
+		}
+	}
+	keys := []string{}
+	for k := range recv { // map[string]*Send
+		keys = append(keys, k)
+	}
+	sort.Strings(keys)
+	//if obj.Debug && len(keys) > 0 {
+	//	// NOTE: this could expose private resource data like passwords
+	//	obj.Logf("SendRecv: %s recv: %+v", res, strings.Join(keys, ", "))
+	//}
+	for k, v := range recv { // map[string]*Send
+		// v.Res // SendableRes // a handle to the resource which is sending a value
+		// v.Key // string      // the key in the resource that we're sending
+		if _, exists := updated[res]; !exists {
+			updated[res] = make(map[string]*engine.Send)
+		}
+
+		//updated[res][k] = false // default
+		v.Changed = false   // reset to the default
+		updated[res][k] = v // default
 
 		var st interface{} = v.Res // old style direct send/recv
 		if true {                  // new style send/recv API
 			st = v.Res.Sent()
 		}
 
+		if st == nil {
+			e := fmt.Errorf("received nil value from: %s", v.Res)
+			err = errwrap.Append(err, e) // list of errors
+			continue
+		}
+
+		if e := engineUtil.StructFieldCompat(st, v.Key, res, k); e != nil {
+			err = errwrap.Append(err, e) // list of errors
+			continue
+		}
+
 		// send
+		m1, e := engineUtil.StructTagToFieldName(st)
+		if e != nil {
+			err = errwrap.Append(err, e) // list of errors
+			continue
+		}
+		key1, exists := m1[v.Key]
+		if !exists {
+			e := fmt.Errorf("requested key of `%s` not found in send struct", v.Key)
+			err = errwrap.Append(err, e) // list of errors
+			continue
+		}
+
 		obj1 := reflect.Indirect(reflect.ValueOf(st))
-		type1 := obj1.Type()
-		value1 := obj1.FieldByName(v.Key)
+		//type1 := obj1.Type()
+		value1 := obj1.FieldByName(key1)
 		kind1 := value1.Kind()
 
 		// recv
+		m2, e := engineUtil.StructTagToFieldName(res)
+		if e != nil {
+			err = errwrap.Append(err, e) // list of errors
+			continue
+		}
+		key2, exists := m2[k]
+		if !exists {
+			e := fmt.Errorf("requested key of `%s` not found in recv struct", k)
+			err = errwrap.Append(err, e) // list of errors
+			continue
+		}
+
 		obj2 := reflect.Indirect(reflect.ValueOf(res)) // pass in full struct
-		type2 := obj2.Type()
-		value2 := obj2.FieldByName(k)
+		//type2 := obj2.Type()
+		value2 := obj2.FieldByName(key2)
 		kind2 := value2.Kind()
 
-		if obj.Debug {
-			obj.Logf("Send(%s) has %v: %v", type1, kind1, value1)
-			obj.Logf("Recv(%s) has %v: %v", type2, kind2, value2)
+		//orig := value1
+		dest := value2 // save the o.g. because we need the real dest!
+
+		// NOTE: Reminder: obj1 comes from st and it is the *<Res>Sends
+		// struct which contains whichever fields that resource sends.
+		// For example, this might be *TestSends for the Test resource.
+		// The receiver is obj2 and that is actually the resource struct
+		// which is a *<Res> and which gets it's fields directly set on.
+		// For example, this might be *TestRes for the Test resource.
+		//fmt.Printf("obj1(%T): %+v\n", obj1, obj1)
+		//fmt.Printf("obj2(%T): %+v\n", obj2, obj2)
+		// Lastly, remember that many of the type incompatibilities are
+		// caught during type unification, and so we might have overly
+		// relaxed the checks here and something could slip by. If we
+		// find something, this code will need new checks added back.
+
+		// Here we unpack one-level, and then leave the complex stuff
+		// for the Into() method below.
+		// for kind1 == reflect.Interface || kind1 == reflect.Ptr // wrong
+		// if kind1 == reflect.Interface || kind1 == reflect.Ptr  // wrong
+		// for kind1 == reflect.Interface // wrong
+		if kind1 == reflect.Interface {
+			value1 = value1.Elem() // un-nest one interface
+			kind1 = value1.Kind()
 		}
 
-		// i think we probably want the same kind, at least for now...
-		if kind1 != kind2 {
-			e := fmt.Errorf("kind mismatch between %s: %s and %s: %s", v.Res, kind1, res, kind2)
-			err = multierr.Append(err, e) // list of errors
-			continue
+		// This second block is identical, but it's just accidentally
+		// symmetrical. The types of input structs are different shapes.
+		// for kind2 == reflect.Interface || kind2 == reflect.Ptr // wrong
+		// if kind2 == reflect.Interface || kind2 == reflect.Ptr  // wrong
+		// for kind2 == reflect.Interface // wrong
+		if kind2 == reflect.Interface {
+			value2 = value2.Elem() // un-nest one interface
+			kind2 = value2.Kind()
 		}
 
-		// if the types don't match, we can't use send->recv
-		// FIXME: do we want to relax this for string -> *string ?
-		if e := TypeCmp(value1, value2); e != nil {
-			e := errwrap.Wrapf(e, "type mismatch between %s and %s", v.Res, res)
-			err = multierr.Append(err, e) // list of errors
-			continue
-		}
+		//if obj.Debug {
+		//	obj.Logf("Send(%s) has %v: %v", type1, kind1, value1)
+		//	obj.Logf("Recv(%s) has %v: %v", type2, kind2, value2)
+		//}
+
+		// Skip this check in favour of the more complex Into() below...
+		//if kind1 != kind2 {
+		//	e := fmt.Errorf("send/recv kind mismatch between %s: %s and %s: %s", v.Res, kind1, res, kind2)
+		//	err = errwrap.Append(err, e) // list of errors
+		//	continue
+		//}
+
+		// Skip this check in favour of the more complex Into() below...
+		//if e := TypeCmp(value1, value2); e != nil {
+		//	e := errwrap.Wrapf(e, "type mismatch between %s and %s", v.Res, res)
+		//	err = errwrap.Append(err, e) // list of errors
+		//	continue
+		//}
 
 		// if we can't set, then well this is pointless!
-		if !value2.CanSet() {
+		if !dest.CanSet() {
 			e := fmt.Errorf("can't set %s.%s", res, k)
-			err = multierr.Append(err, e) // list of errors
+			err = errwrap.Append(err, e) // list of errors
 			continue
 		}
 
 		// if we can't interface, we can't compare...
-		if !value1.CanInterface() || !value2.CanInterface() {
+		if !value1.CanInterface() {
+			e := fmt.Errorf("can't interface %s.%s", v.Res, v.Key)
+			err = errwrap.Append(err, e) // list of errors
+			continue
+		}
+		if !value2.CanInterface() {
 			e := fmt.Errorf("can't interface %s.%s", res, k)
-			err = multierr.Append(err, e) // list of errors
+			err = errwrap.Append(err, e) // list of errors
 			continue
 		}
 
 		// if the values aren't equal, we're changing the receiver
-		if !reflect.DeepEqual(value1.Interface(), value2.Interface()) {
-			// TODO: can we catch the panics here in case they happen?
-			value2.Set(value1) // do it for all types that match
-			updated[k] = true  // we updated this key!
-			v.Changed = true   // tag this key as updated!
-			obj.Logf("SendRecv: %s.%s -> %s.%s", v.Res, v.Key, res, k)
+		if reflect.DeepEqual(value1.Interface(), value2.Interface()) {
+			continue // skip as they're the same, no error needed
 		}
+
+		// TODO: can we catch the panics here in case they happen?
+
+		fv, e := types.ValueOf(value1)
+		if e != nil {
+			e := errwrap.Wrapf(e, "bad value %s.%s", v.Res, v.Key)
+			err = errwrap.Append(err, e) // list of errors
+			continue
+		}
+
+		// mutate the struct field dest with the mcl data in fv
+		if e := types.Into(fv, dest); e != nil {
+			// runtime error, probably from using value res
+			e := errwrap.Wrapf(e, "mismatch: %s.%s (%s) -> %s.%s (%s)", v.Res, v.Key, kind1, res, k, kind2)
+			err = errwrap.Append(err, e) // list of errors
+			continue
+		}
+		//dest.Set(orig)  // do it for all types that match
+		//updated[res][k] = true // we updated this key!
+		v.Changed = true    // tag this key as updated!
+		updated[res][k] = v // we updated this key!
+		//obj.Logf("SendRecv: %s.%s -> %s.%s (%+v)", v.Res, v.Key, res, k, fv) // fv may be private data
 	}
 	return updated, err
 }
@@ -116,3 +289,19 @@ func TypeCmp(a, b reflect.Value) error {
 
 	return nil // identical Type()'s
 }
+
+// UpdatedStrings returns a list of strings showing what was updated after a
+// Send/Recv run returned the updated datastructure. This is useful for logs.
+func UpdatedStrings(updated map[engine.RecvableRes]map[string]*engine.Send) []string {
+	out := []string{}
+	for r, m := range updated { // map[engine.RecvableRes]map[string]*engine.Send
+		for s, send := range m {
+			if !send.Changed {
+				continue
+			}
+			x := fmt.Sprintf("%v.%s -> %v.%s", send.Res, send.Key, r, s)
+			out = append(out, x)
+		}
+	}
+	return out
+}

engine/graph/state.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,30 +13,41 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package graph
 
 import (
+	"context"
 	"fmt"
-	"os"
-	"path"
 	"sync"
+	"sync/atomic"
 	"time"
 
 	"github.com/purpleidea/mgmt/converger"
 	"github.com/purpleidea/mgmt/engine"
-	"github.com/purpleidea/mgmt/engine/event"
+	"github.com/purpleidea/mgmt/engine/local"
+	engineUtil "github.com/purpleidea/mgmt/engine/util"
 	"github.com/purpleidea/mgmt/pgraph"
-	"github.com/purpleidea/mgmt/util"
-
-	errwrap "github.com/pkg/errors"
+	"github.com/purpleidea/mgmt/util/errwrap"
 )
 
 // State stores some state about the resource it is mapped to.
 type State struct {
 	// Graph is a pointer to the graph that this vertex is part of.
-	//Graph pgraph.Graph
+	Graph *pgraph.Graph
 
 	// Vertex is the pointer in the graph that this state corresponds to. It
 	// can be converted to a `Res` if necessary.
@@ -44,65 +55,92 @@ type State struct {
 	Vertex pgraph.Vertex
 
 	Program  string
+	Version  string
 	Hostname string
-	World    engine.World
+
+	//Converger *converger.Coordinator
+
+	Local *local.API
+	World engine.World
 
 	// Prefix is a unique directory prefix which can be used. It should be
 	// created if needed.
 	Prefix string
 
-	//Converger converger.Converger
-
 	// Debug turns on additional output and behaviours.
 	Debug bool
 
 	// Logf is the logging function that should be used to display messages.
 	Logf func(format string, v ...interface{})
 
-	timestamp int64 // last updated timestamp
-	isStateOK bool  // is state OK or do we need to run CheckApply ?
+	timestamp int64        // last updated timestamp
+	isStateOK *atomic.Bool // is state OK or do we need to run CheckApply ?
+	workerErr error        // did the Worker error?
 
-	// events is a channel of incoming events which is read by the Watch
-	// loop for that resource. It receives events like pause, start, and
-	// poke. The channel shuts down to signal for Watch to exit.
-	eventsChan chan event.Kind // incoming to resource
-	eventsLock *sync.Mutex     // lock around sending and closing of events channel
-	eventsDone bool            // is channel closed?
+	mutex *sync.RWMutex // used for editing state properties
 
-	// outputChan is the channel that the engine listens on for events from
+	// doneCtx is cancelled when Watch should shut down. When any of the
+	// following channels close, it causes this to close.
+	doneCtx context.Context
+
+	// doneCtxCancel is the cancel function for doneCtx.
+	doneCtxCancel func()
+
+	// processDone is closed when the Process/CheckApply function fails
+	// permanently, and wants to cause Watch to exit.
+	processDone chan struct{}
+	// watchDone is closed when the Watch function fails permanently, and we
+	// close this to signal we should definitely exit. (Often redundant.)
+	watchDone chan struct{} // could be shared with limitDone or retryDone
+	// limitDone is closed when the Watch function fails permanently, and we
+	// close this to signal we should definitely exit. This happens inside
+	// of the limit loop of the Process section of Worker.
+	limitDone chan struct{} // could be shared with watchDone or retryDone
+	// retryDone is closed when the Watch function fails permanently, and we
+	// close this to signal we should definitely exit. This happens inside
+	// of the retry loop of the Process section of Worker.
+	retryDone chan struct{} // could be shared with watchDone or limitDone
+	// removeDone is closed when the vertexRemoveFn method asks for an exit.
+	// This happens when we're switching graphs. The switch to an "empty" is
+	// the equivalent of asking for a final shutdown.
+	removeDone chan struct{}
+	// eventsDone is closed when we shutdown the Process loop because we
+	// closed without error. In theory this shouldn't happen, but it could
+	// if Watch returns without error for some reason.
+	eventsDone chan struct{}
+
+	// eventsChan is the channel that the engine listens on for events from
 	// the Watch loop for that resource. The event is nil normally, except
 	// when events are sent on this channel from the engine. This only
 	// happens as a signaling mechanism when Watch has shutdown and we want
 	// to notify the Process loop which reads from this.
-	outputChan chan error // outgoing from resource
+	eventsChan chan error // outgoing from resource
 
-	wg   *sync.WaitGroup
-	exit *util.EasyExit
+	// pokeChan is a separate channel that the Process loop listens on to
+	// know when we might need to run Process. It never closes, and is safe
+	// to send on since it is buffered.
+	pokeChan chan struct{} // outgoing from resource
 
-	started chan struct{} // closes when it's started
-	stopped chan struct{} // closes when it's stopped
+	// paused represents if this particular res is paused or not. This is
+	// primarily used to avoid running an unnecessary Resume on the first
+	// run of this resource.
+	paused bool
+	// pauseSignal receives a message to request a pause of this resource.
+	pauseSignal chan struct{}
+	// resumeSignal receives a message to resume this resource. The channel
+	// closes when the resource is removed from the graph.
+	resumeSignal chan struct{}
 
-	starter bool // do we have an indegree of 0 ?
-	working bool // is the Main() loop running ?
+	wg *sync.WaitGroup // used for all vertex specific processes
 
-	cuid converger.UID // primary converger
+	cuid *converger.UID // primary converger
+	tuid *converger.UID // secondary converger
 
 	init *engine.Init // a copy of the init struct passed to res Init
 }
 
 // Init initializes structures like channels.
 func (obj *State) Init() error {
-	obj.eventsChan = make(chan event.Kind)
-	obj.eventsLock = &sync.Mutex{}
-
-	obj.outputChan = make(chan error)
-
-	obj.wg = &sync.WaitGroup{}
-	obj.exit = util.NewEasyExit()
-
-	obj.started = make(chan struct{})
-	obj.stopped = make(chan struct{})
-
 	res, isRes := obj.Vertex.(engine.Res)
 	if !isRes {
 		return fmt.Errorf("vertex is not a Res")
@@ -120,29 +158,39 @@ func (obj *State) Init() error {
 		return fmt.Errorf("the Logf function is missing")
 	}
 
+	obj.isStateOK = &atomic.Bool{}
+
+	obj.mutex = &sync.RWMutex{}
+	obj.doneCtx, obj.doneCtxCancel = context.WithCancel(context.Background())
+
+	obj.processDone = make(chan struct{})
+	obj.watchDone = make(chan struct{})
+	obj.limitDone = make(chan struct{})
+	obj.retryDone = make(chan struct{})
+	obj.removeDone = make(chan struct{})
+	obj.eventsDone = make(chan struct{})
+
+	obj.eventsChan = make(chan error)
+
+	obj.pokeChan = make(chan struct{}, 1) // must be buffered
+
+	//obj.paused = false // starts off as started
+	obj.pauseSignal = make(chan struct{})
+	obj.resumeSignal = make(chan struct{})
+
+	obj.wg = &sync.WaitGroup{}
+
 	//obj.cuid = obj.Converger.Register() // gets registered in Worker()
+	//obj.tuid = obj.Converger.Register() // gets registered in Worker()
 
 	obj.init = &engine.Init{
 		Program:  obj.Program,
+		Version:  obj.Version,
 		Hostname: obj.Hostname,
 
 		// Watch:
-		Running: func() error {
-			close(obj.started)    // this is reset in the reset func
-			obj.isStateOK = false // assume we're initially dirty
-			// optimization: skip the initial send if not a starter
-			// because we'll get poked from a starter soon anyways!
-			if !obj.starter {
-				return nil
-			}
-			return obj.event()
-		},
-		Event:  obj.event,
-		Events: obj.eventsChan,
-		Read:   obj.read,
-		Dirty: func() { // TODO: should we rename this SetDirty?
-			obj.isStateOK = false
-		},
+		Running: obj.event,
+		Event:   obj.event,
 
 		// CheckApply:
 		Refresh: func() bool {
@@ -152,33 +200,72 @@ func (obj *State) Init() error {
 			}
 			return res.Refresh()
 		},
-		Send: func(st interface{}) error {
-			res, ok := obj.Vertex.(engine.SendableRes)
-			if !ok {
-				panic("res does not support the Sendable trait")
-			}
-			// XXX: type check this
-			//expected := res.Sends()
-			//if err := XXX_TYPE_CHECK(expected, st); err != nil {
-			//	return err
-			//}
 
-			return res.Send(st) // send the struct
-		},
-		Recv: func() map[string]*engine.Send { // TODO: change this API?
-			res, ok := obj.Vertex.(engine.RecvableRes)
-			if !ok {
-				panic("res does not support the Recvable trait")
+		Send: engine.GenerateSendFunc(res),
+		Recv: engine.GenerateRecvFunc(res),
+
+		// FIXME: pass in a safe, limited query func instead?
+		// TODO: not implemented, use FilteredGraph
+		//Graph: func() *pgraph.Graph {
+		//	_, ok := obj.Vertex.(engine.CanGraphQueryRes)
+		//	if !ok {
+		//		panic("res does not support the GraphQuery trait")
+		//	}
+		//	return obj.Graph // we return in a func so it's fresh!
+		//},
+
+		FilteredGraph: func() (*pgraph.Graph, error) {
+			graph, err := pgraph.NewGraph("filtered")
+			if err != nil {
+				return nil, err
 			}
-			return res.Recv()
+
+			// filter graph and build a new one...
+			adjacency := obj.Graph.Adjacency()
+			for v1 := range adjacency {
+				// check we're allowed
+				r1, ok := v1.(engine.GraphQueryableRes)
+				if !ok {
+					continue
+				}
+				// pass in information on requestor...
+				if err := r1.GraphQueryAllowed(
+					engine.GraphQueryableOptionKind(res.Kind()),
+					engine.GraphQueryableOptionName(res.Name()),
+					// TODO: add more information...
+				); err != nil {
+					continue
+				}
+				graph.AddVertex(v1)
+
+				for v2, edge := range adjacency[v1] {
+					r2, ok := v2.(engine.GraphQueryableRes)
+					if !ok {
+						continue
+					}
+					// pass in information on requestor...
+					if err := r2.GraphQueryAllowed(
+						engine.GraphQueryableOptionKind(res.Kind()),
+						engine.GraphQueryableOptionName(res.Name()),
+						// TODO: add more information...
+					); err != nil {
+						continue
+					}
+					//graph.AddVertex(v2) // redundant
+					graph.AddEdge(v1, v2, edge)
+				}
+			}
+
+			return graph, nil // we return in a func so it's fresh!
 		},
 
+		Local:  obj.Local,
 		World:  obj.World,
 		VarDir: obj.varDir,
 
 		Debug: obj.Debug,
 		Logf: func(format string, v ...interface{}) {
-			obj.Logf("resource: "+format, v...)
+			obj.Logf(format, v...)
 		},
 	}
 
@@ -186,9 +273,15 @@ func (obj *State) Init() error {
 	if obj.Debug {
 		obj.Logf("Init(%s)", res)
 	}
+
+	// write the reverse request to the disk...
+	if err := obj.ReversalInit(); err != nil {
+		return err // TODO: test this code path...
+	}
+
 	err := res.Init(obj.init)
 	if obj.Debug {
-		obj.Logf("Init(%s): Return(%+v)", res, err)
+		obj.Logf("Init(%s): Return(%s)", res, engineUtil.CleanError(err))
 	}
 	if err != nil {
 		return errwrap.Wrapf(err, "could not Init() resource")
@@ -197,9 +290,9 @@ func (obj *State) Init() error {
 	return nil
 }
 
-// Close shuts down and performs any cleanup. This is most akin to a "post" or
+// Cleanup shuts down and performs any cleanup. This is most akin to a "post" or
 // cleanup command as the initiator for closing a vertex happens in graph sync.
-func (obj *State) Close() error {
+func (obj *State) Cleanup() error {
 	res, isRes := obj.Vertex.(engine.Res)
 	if !isRes {
 		return fmt.Errorf("vertex is not a Res")
@@ -208,6 +301,9 @@ func (obj *State) Close() error {
 	//if obj.cuid != nil {
 	//	obj.cuid.Unregister() // gets unregistered in Worker()
 	//}
+	//if obj.tuid != nil {
+	//	obj.tuid.Unregister() // gets unregistered in Worker()
+	//}
 
 	// redundant safety
 	obj.wg.Wait() // wait until all poke's and events on me have exited
@@ -216,221 +312,121 @@ func (obj *State) Close() error {
 	if obj.Debug {
 		obj.Logf("Close(%s)", res)
 	}
-	err := res.Close()
+
+	var reverr error
+	// clear the reverse request from the disk...
+	if err := obj.ReversalCleanup(); err != nil {
+		// TODO: test this code path...
+		// TODO: should this be an error or a warning?
+		reverr = err
+	}
+
+	reterr := res.Cleanup()
 	if obj.Debug {
-		obj.Logf("Close(%s): Return(%+v)", res, err)
+		obj.Logf("Close(%s): Return(%s)", res, engineUtil.CleanError(reterr))
 	}
 
-	return err
+	reterr = errwrap.Append(reterr, reverr)
+
+	return reterr
 }
 
-// reset is run to reset the state so that Watch can run a second time. Thus is
-// needed for the Watch retry in particular.
-func (obj *State) reset() {
-	obj.started = make(chan struct{})
-	obj.stopped = make(chan struct{})
-}
-
-// Poke sends a nil message on the outputChan. This channel is used by the
-// resource to signal a possible change. This will cause the Process loop to
-// run if it can.
+// Poke sends a notification on the poke channel. This channel is used to notify
+// the Worker to run the Process/CheckApply when it can. This is used when there
+// is a need to schedule or reschedule some work which got postponed or dropped.
+// This doesn't contain any internal synchronization primitives or wait groups,
+// callers are expected to make sure that they don't leave any of these running
+// by the time the Worker() shuts down.
 func (obj *State) Poke() {
-	// add a wait group on the vertex we're poking!
-	obj.wg.Add(1)
-	defer obj.wg.Done()
-
 	select {
-	case obj.outputChan <- nil:
-
-	case <-obj.exit.Signal():
+	case obj.pokeChan <- struct{}{}:
+	default: // if chan is now full because more than one poke happened...
 	}
 }
 
-// Event sends a Pause or Start event to the resource. It can also be used to
-// send Poke events, but it's much more efficient to send them directly instead
-// of passing them through the resource.
-func (obj *State) Event(kind event.Kind) {
-	// TODO: should these happen after the lock?
-	obj.wg.Add(1)
-	defer obj.wg.Done()
-
-	obj.eventsLock.Lock()
-	defer obj.eventsLock.Unlock()
-
-	if obj.eventsDone { // closing, skip events...
-		return
+// Pause pauses this resource. It must not be called on any already paused
+// resource. It will block until the resource pauses with an acknowledgment, or
+// until an exit for that resource is seen. If the latter happens it will error.
+// It must not be called concurrently with either the Resume() method or itself,
+// so only call these one at a time and alternate between the two.
+func (obj *State) Pause() error {
+	if obj.paused {
+		panic("already paused")
 	}
 
-	if kind == event.EventExit { // set this so future events don't deadlock
-		obj.Logf("exit event...")
-		obj.eventsDone = true
-		close(obj.eventsChan) // causes resource Watch loop to close
-		obj.exit.Done(nil)    // trigger exit signal to unblock some cases
+	// wait for ack (or exit signal)
+	select {
+	case obj.pauseSignal <- struct{}{}:
+		// we're paused
+
+	case <-obj.doneCtx.Done():
+		return engine.ErrClosed
+	}
+	obj.paused = true
+
+	return nil
+}
+
+// Resume unpauses this resource. It can be safely called once on a brand-new
+// resource that has just started running, without incident. It must not be
+// called concurrently with either the Pause() method or itself, so only call
+// these one at a time and alternate between the two.
+func (obj *State) Resume() {
+	// This paused check prevents unnecessary "resume" calls to the resource
+	// on its first run, since resources start in the running state!
+	if !obj.paused { // no need to unpause brand-new resources
 		return
 	}
 
 	select {
-	case obj.eventsChan <- kind:
-
-	case <-obj.exit.Signal():
+	case obj.resumeSignal <- struct{}{}:
 	}
+
+	obj.paused = false
 }
 
-// read is a helper function used inside the main select statement of resources.
-// If it returns an error, then this is a signal for the resource to exit.
-func (obj *State) read(kind event.Kind) error {
-	switch kind {
-	case event.EventPoke:
-		return obj.event() // a poke needs to cause an event...
-	case event.EventStart:
-		return fmt.Errorf("unexpected start")
-	case event.EventPause:
-		// pass
-	case event.EventExit:
-		return engine.ErrSignalExit
+// event is a helper function to send an event to the CheckApply process loop.
+// It can be used for the initial `running` event, or any regular event. You
+// should instead use Poke() to "schedule" a new Process/CheckApply loop when
+// one might be needed. This method will block until we're unpaused and ready to
+// receive on the events channel.
+func (obj *State) event() {
+	obj.setDirty() // assume we're initially dirty
 
-	default:
-		return fmt.Errorf("unhandled event: %+v", kind)
-	}
-
-	// we're paused now
 	select {
-	case kind, ok := <-obj.eventsChan:
-		if !ok {
-			return engine.ErrWatchExit
-		}
-		switch kind {
-		case event.EventPoke:
-			return fmt.Errorf("unexpected poke")
-		case event.EventPause:
-			return fmt.Errorf("unexpected pause")
-		case event.EventStart:
-			// resumed
-			return nil
-		case event.EventExit:
-			return engine.ErrSignalExit
-
-		default:
-			return fmt.Errorf("unhandled event: %+v", kind)
-		}
+	case obj.eventsChan <- nil: // blocks! (this is unbuffered)
+		// send!
 	}
+
+	//return // implied
 }
 
-// event is a helper function to send an event from the resource Watch loop. It
-// can be used for the initial `running` event, or any regular event. If it
-// returns an error, then the Watch loop must return this error and shutdown.
-func (obj *State) event() error {
-	// loop until we sent on obj.outputChan or exit with error
-	for {
-		select {
-		// send "activity" event
-		case obj.outputChan <- nil:
-			return nil // sent event!
-
-		// make sure to keep handling incoming
-		case kind, ok := <-obj.eventsChan:
-			if !ok {
-				return engine.ErrWatchExit
-			}
-			switch kind {
-			case event.EventPoke:
-				// we're trying to send an event, so swallow the
-				// poke: it's what we wanted to have happen here
-				continue
-			case event.EventStart:
-				return fmt.Errorf("unexpected start")
-			case event.EventPause:
-				// pass
-			case event.EventExit:
-				return engine.ErrSignalExit
-
-			default:
-				return fmt.Errorf("unhandled event: %+v", kind)
-			}
-		}
-
-		// we're paused now
-		select {
-		case kind, ok := <-obj.eventsChan:
-			if !ok {
-				return engine.ErrWatchExit
-			}
-			switch kind {
-			case event.EventPoke:
-				return fmt.Errorf("unexpected poke")
-			case event.EventPause:
-				return fmt.Errorf("unexpected pause")
-			case event.EventStart:
-				// resumed
-			case event.EventExit:
-				return engine.ErrSignalExit
-
-			default:
-				return fmt.Errorf("unhandled event: %+v", kind)
-			}
-		}
-	}
-}
-
-// varDir returns the path to a working directory for the resource. It will try
-// and create the directory first, and return an error if this failed. The dir
-// should be cleaned up by the resource on Close if it wishes to discard the
-// contents. If it does not, then a future resource with the same kind and name
-// may see those contents in that directory. The resource should clean up the
-// contents before use if it is important that nothing exist. It is always
-// possible that contents could remain after an abrupt crash, so do not store
-// overly sensitive data unless you're aware of the risks.
-func (obj *State) varDir(extra string) (string, error) {
-	// Using extra adds additional dirs onto our namespace. An empty extra
-	// adds no additional directories.
-	if obj.Prefix == "" { // safety
-		return "", fmt.Errorf("the VarDir prefix is empty")
-	}
-
-	// an empty string at the end has no effect
-	p := fmt.Sprintf("%s/", path.Join(obj.Prefix, extra))
-	if err := os.MkdirAll(p, 0770); err != nil {
-		return "", errwrap.Wrapf(err, "can't create prefix in: %s", p)
-	}
-
-	// returns with a trailing slash as per the mgmt file res convention
-	return p, nil
+// setDirty marks the resource state as dirty. This signals to the engine that
+// CheckApply will have some work to do in order to converge it.
+func (obj *State) setDirty() {
+	obj.tuid.StopTimer()
+	//obj.mutex.Lock()
+	obj.isStateOK.Store(false) // concurrent write
+	//obj.mutex.Unlock()
 }
 
 // poll is a replacement for Watch when the Poll metaparameter is used.
-func (obj *State) poll(interval uint32) error {
+func (obj *State) poll(ctx context.Context, interval uint32) error {
 	// create a time.Ticker for the given interval
 	ticker := time.NewTicker(time.Duration(interval) * time.Second)
 	defer ticker.Stop()
 
-	// notify engine that we're running
-	if err := obj.init.Running(); err != nil {
-		return err // exit if requested
-	}
+	obj.init.Running() // when started, notify engine that we're running
 
-	var send = false // send event?
 	for {
 		select {
 		case <-ticker.C: // received the timer event
 			obj.init.Logf("polling...")
-			send = true
-			obj.init.Dirty() // dirty
 
-		case event, ok := <-obj.init.Events:
-			if !ok {
-				return nil
-			}
-			if err := obj.init.Read(event); err != nil {
-				return err
-			}
+		case <-ctx.Done(): // signal for shutdown request
+			return nil
 		}
 
-		// do all our event sending all together to avoid duplicate msgs
-		if send {
-			send = false
-			if err := obj.init.Event(); err != nil {
-				return err // exit if requested
-			}
-		}
+		obj.init.Event() // notify engine of an event (this can block)
 	}
 }

engine/graph/vardir.go

@@ -0,0 +1,63 @@
+// Mgmt
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
+
+package graph
+
+import (
+	"fmt"
+	"os"
+	"path"
+
+	"github.com/purpleidea/mgmt/util/errwrap"
+)
+
+// varDir returns the path to a working directory for the resource. It will try
+// and create the directory first, and return an error if this failed. The dir
+// should be cleaned up by the resource on Close if it wishes to discard the
+// contents. If it does not, then a future resource with the same kind and name
+// may see those contents in that directory. The resource should clean up the
+// contents before use if it is important that nothing exist. It is always
+// possible that contents could remain after an abrupt crash, so do not store
+// overly sensitive data unless you're aware of the risks.
+func (obj *State) varDir(extra string) (string, error) {
+	// Using extra adds additional dirs onto our namespace. An empty extra
+	// adds no additional directories.
+	if obj.Prefix == "" { // safety
+		return "", fmt.Errorf("the VarDir prefix is empty")
+	}
+
+	// an empty string at the end has no effect
+	p := fmt.Sprintf("%s/", path.Join(obj.Prefix, extra))
+	if err := os.MkdirAll(p, 0770); err != nil {
+		return "", errwrap.Wrapf(err, "can't create prefix in: %s", p)
+	}
+
+	// returns with a trailing slash as per the mgmt file res convention
+	return p, nil
+}

engine/graphqueryable.go

@@ -0,0 +1,82 @@
+// Mgmt
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
+
+package engine
+
+// GraphQueryableRes is the interface that must be implemented if you want your
+// resource to be allowed to be queried from another resource in the graph. This
+// is done as a form of explicit authorization tracking so that we can consider
+// security aspects more easily. Ultimately, all resource code should be
+// trusted, but it's still a good idea to know if a particular resource is even
+// able to access information about another one, and if your resource doesn't
+// add the trait supporting this, then it won't be allowed.
+type GraphQueryableRes interface {
+	Res // implement everything in Res but add the additional requirements
+
+	// GraphQueryAllowed returns nil if you're allowed to query the graph.
+	GraphQueryAllowed(...GraphQueryableOption) error
+}
+
+// GraphQueryableOption is an option that can be used to specify the
+// authentication.
+type GraphQueryableOption func(*GraphQueryableOptions)
+
+// GraphQueryableOptions represents the different possible configurable options.
+type GraphQueryableOptions struct {
+	// Kind is the kind of the resource making the access.
+	Kind string
+	// Name is the name of the resource making the access.
+	Name string
+	// TODO: add more options if needed
+}
+
+// Apply is a helper function to apply a list of options to the struct. You
+// should initialize it with defaults you want, and then apply any you've
+// received like this.
+func (obj *GraphQueryableOptions) Apply(opts ...GraphQueryableOption) {
+	for _, optionFunc := range opts { // apply the options
+		optionFunc(obj)
+	}
+}
+
+// GraphQueryableOptionKind tells the GraphQueryAllowed function what the
+// resource kind is.
+func GraphQueryableOptionKind(kind string) GraphQueryableOption {
+	return func(gqo *GraphQueryableOptions) {
+		gqo.Kind = kind
+	}
+}
+
+// GraphQueryableOptionName tells the GraphQueryAllowed function what the
+// resource name is.
+func GraphQueryableOptionName(name string) GraphQueryableOption {
+	return func(gqo *GraphQueryableOptions) {
+		gqo.Name = name
+	}
+}

engine/local/local.go

@@ -0,0 +1,334 @@
+// Mgmt
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
+
+// Package local contains functions and interfaces that are shared between
+// functions and resources. It's similar to the "world" functionality, except
+// that it only involves local operations that stay within a single machine or
+// local mgmt instance.
+package local
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"path"
+	"strings"
+	"sync"
+
+	"github.com/purpleidea/mgmt/util"
+)
+
+// API implements the base handle for all the methods in this package. If we
+// were going to have more than one implementation for all of these, then this
+// would be an interface instead, and different packages would implement it.
+// Since this is not the expectation for the local API, it's all self-contained.
+type API struct {
+	Prefix string
+	Debug  bool
+	Logf   func(format string, v ...interface{})
+
+	// Each piece of the API can take a handle here.
+	*Value
+}
+
+// Init initializes the API before first use. It returns itself so it can be
+// chained for API aesthetical purposes.
+func (obj *API) Init() *API {
+	obj.Value = &Value{}
+	obj.Value.Init(&ValueInit{
+		Prefix: obj.Prefix,
+		Debug:  obj.Debug,
+		Logf:   obj.Logf,
+	})
+
+	return obj
+}
+
+// ValueInit are the init values that the Value API needs to work correctly.
+type ValueInit struct {
+	Prefix string
+	Debug  bool
+	Logf   func(format string, v ...interface{})
+}
+
+// Value is the API for getting, setting, and watching local values.
+type Value struct {
+	init         *ValueInit
+	mutex        *sync.Mutex
+	prefix       string
+	prefixExists bool // is it okay to use the prefix?
+	values       map[string]interface{}
+	notify       map[chan struct{}]string // one chan (unique ptr) for each watch
+	skipread     map[string]struct{}
+}
+
+// Init runs some initialization code for the Value API.
+func (obj *Value) Init(init *ValueInit) {
+	obj.init = init
+	obj.mutex = &sync.Mutex{}
+	obj.prefix = fmt.Sprintf("%s/", path.Join(obj.init.Prefix, "value"))
+	obj.values = make(map[string]interface{})
+	obj.notify = make(map[chan struct{}]string)
+	obj.skipread = make(map[string]struct{})
+
+	// We don't need to, or want to, load any of the keys from disk
+	// initially, because (1) this would consume memory for keys we never
+	// use, and (2) we can load them on first read instead.
+	// TODO: build in some sort of expiry system that deletes keys older
+	// than X weeks to prevent infinite growth of the on-disk database.
+}
+
+// ValueGet pulls a value out of a local in-memory, key-value store that is
+// backed by on-disk storage. While each value is intended to have an underlying
+// type, we use the `any` or empty `interface{}` value to represent each value
+// instead of a `types.Value` because it's more generic, and not limited to
+// being used with the language type system. If the value doesn't exist, we
+// return a nil value and no error.
+func (obj *Value) ValueGet(ctx context.Context, key string) (interface{}, error) {
+	prefix, err := obj.getPrefix()
+	if err != nil {
+		return nil, err
+	}
+
+	obj.mutex.Lock()
+	defer obj.mutex.Unlock()
+
+	var val interface{}
+	//var err error
+	if _, skip := obj.skipread[key]; skip {
+		val, err = valueRead(ctx, prefix, key) // must return val == nil if missing
+		if err != nil {
+			// We had an actual read issue. Report this and stop
+			// because it means we might not be allowing our
+			// cold-cache warming if we ignored it.
+			return nil, err
+		}
+		// File not found errors are masked in the valueRead function
+	}
+
+	// Anything in memory, will override whatever we might have read.
+	value, exists := obj.values[key]
+	if !exists {
+		// disable future disk reads since the cache is now warm!
+		obj.skipread[key] = struct{}{}
+		return val, nil // if val is nil, we didn't find it
+	}
+	return value, nil
+}
+
+// ValueSet sets a value to our in-memory, key-value store that is backed by
+// on-disk storage. If you provide a nil value, this is the equivalent of
+// removing or deleting the value.
+func (obj *Value) ValueSet(ctx context.Context, key string, value interface{}) error {
+	prefix, err := obj.getPrefix()
+	if err != nil {
+		return err
+	}
+
+	obj.mutex.Lock()
+	defer obj.mutex.Unlock()
+
+	// Write to state dir on disk first. If ctx cancels, we assume it's not
+	// written or it doesn't matter because we're cancelling, meaning we're
+	// shutting down, so our local cache can be invalidated anyways.
+
+	if value == nil { // remove/delete
+		if err := valueRemove(ctx, prefix, key); err != nil {
+			return err
+		}
+	} else {
+		if err := valueWrite(ctx, prefix, key, value); err != nil {
+			return err
+		}
+	}
+
+	if value == nil { // remove/delete
+		delete(obj.values, key)
+	} else {
+		obj.values[key] = value // store to in-memory map
+	}
+
+	// We still notify on remove/delete!
+	for ch, k := range obj.notify { // send notifications to any watchers...
+		if k != key { // there might be more than one watcher per key
+			continue
+		}
+		select {
+		case ch <- struct{}{}: // must be async and not block forever
+			// send
+
+			// We don't ever exit here, because that would be the equivalent
+			// of dropping a notification on the floor. This loop is
+			// non-blocking, and so it's okay to just finish it up quickly.
+			//case <-ctx.Done():
+		}
+	}
+
+	return nil
+}
+
+// ValueWatch watches a value from our in-memory, key-value store that is backed
+// by on-disk storage. Conveniently, it never has to watch the on-disk storage,
+// because after the initial startup which always sends a single startup event,
+// it suffices to watch the in-memory store for events!
+func (obj *Value) ValueWatch(ctx context.Context, key string) (chan struct{}, error) {
+	// No need to look at the prefix on disk, because we can do all our
+	// watches from memory!
+	//prefix, err := obj.getPrefix()
+	//if err != nil {
+	//	return nil, err
+	//}
+
+	obj.mutex.Lock()
+	defer obj.mutex.Unlock()
+
+	notifyCh := make(chan struct{}, 1) // so we can async send
+	obj.notify[notifyCh] = key         // add (while within the mutex)
+	notifyCh <- struct{}{}             // startup signal, send one!
+	ch := make(chan struct{})
+	go func() {
+		defer func() { // cleanup
+			obj.mutex.Lock()
+			defer obj.mutex.Unlock()
+			delete(obj.notify, notifyCh) // free memory (in mutex)
+		}()
+		for {
+			select {
+			case _, ok := <-notifyCh:
+				if !ok {
+					// programming error
+					panic("unexpected channel closure")
+				}
+				// recv
+
+			case <-ctx.Done():
+				break // we exit
+			}
+
+			select {
+			case ch <- struct{}{}:
+				// send
+
+			case <-ctx.Done():
+				break // we exit
+			}
+		}
+	}()
+
+	return ch, nil
+}
+
+// getPrefix gets the prefix dir to use, or errors if it can't make one. It
+// makes it on first use, and returns quickly from any future calls to it.
+func (obj *Value) getPrefix() (string, error) {
+	// NOTE: Moving this mutex to just below the first early return, would
+	// be a benign race, but as it turns out, it's possible that a compiler
+	// would see this behaviour as "undefined" and things might not work as
+	// intended. It could perhaps be replaced with a sync/atomic primitive
+	// if we wanted better performance here.
+	obj.mutex.Lock()
+	defer obj.mutex.Unlock()
+
+	if obj.prefixExists { // former race read
+		return obj.prefix, nil
+	}
+
+	// MkdirAll instead of Mkdir because we have no idea if the parent
+	// local/ directory was already made yet or not. (If at all.) If path is
+	// already a directory, MkdirAll does nothing and returns nil. (Good!)
+	// TODO: I hope MkdirAll is thread-safe on path creation in case another
+	// future local API tries to make the base (parent) directory too!
+	if err := os.MkdirAll(obj.prefix, 0755); err != nil {
+		return "", err
+	}
+	obj.prefixExists = true // former race write
+
+	return obj.prefix, nil
+}
+
+func valueRead(ctx context.Context, prefix, key string) (interface{}, error) {
+	// TODO: implement ctx cancellation
+	// TODO: replace with my path library
+	if !strings.HasSuffix(prefix, "/") {
+		return nil, fmt.Errorf("prefix is not a dir")
+	}
+	if !strings.HasPrefix(prefix, "/") {
+		return nil, fmt.Errorf("prefix is not absolute")
+	}
+	p := fmt.Sprintf("%s%s", prefix, key)
+
+	b, err := os.ReadFile(p)
+	if os.IsNotExist(err) {
+		return nil, nil // not found
+	}
+	if err != nil {
+		return nil, err
+	}
+	// file exists!
+	s := string(b)
+	s = strings.TrimSpace(s) // get rid of any newline
+
+	return util.B64ToValue(s)
+}
+
+func valueWrite(ctx context.Context, prefix, key string, value interface{}) error {
+	// TODO: implement ctx cancellation
+	// TODO: replace with my path library
+	if !strings.HasSuffix(prefix, "/") {
+		return fmt.Errorf("prefix is not a dir")
+	}
+	if !strings.HasPrefix(prefix, "/") {
+		return fmt.Errorf("prefix is not absolute")
+	}
+	p := fmt.Sprintf("%s%s", prefix, key)
+
+	s, err := util.ValueToB64(value)
+	if err != nil {
+		return err
+	}
+	s += "\n" // files end with a newline
+	return os.WriteFile(p, []byte(s), 0600)
+}
+
+func valueRemove(ctx context.Context, prefix, key string) error {
+	// TODO: implement ctx cancellation
+	// TODO: replace with my path library
+	if !strings.HasSuffix(prefix, "/") {
+		return fmt.Errorf("prefix is not a dir")
+	}
+	if !strings.HasPrefix(prefix, "/") {
+		return fmt.Errorf("prefix is not absolute")
+	}
+	p := fmt.Sprintf("%s%s", prefix, key)
+
+	if err := os.Remove(p); err != nil && !os.IsNotExist(err) {
+		return err
+	}
+	return nil // ignore not found errors
+}

engine/local/local_test.go

@@ -0,0 +1,62 @@
+// Mgmt
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
+
+//go:build !root
+
+package local
+
+import (
+	"context"
+	"fmt"
+	"reflect"
+	"testing"
+)
+
+func TestWrite(t *testing.T) {
+	tmpdir := fmt.Sprintf("%s/", t.TempDir()) // gets cleaned up at end, new dir for each call
+	key := "test1"
+	value := 42
+	if err := valueWrite(context.Background(), tmpdir, key, value); err != nil {
+		t.Errorf("error: %+v", err)
+		return
+	}
+
+	if val, err := valueRead(context.Background(), tmpdir, key); err != nil {
+		t.Errorf("error: %+v", err)
+		return
+	} else if !reflect.DeepEqual(value, val) {
+		t.Errorf("error: not equal: %+v != %+v", val, value)
+		//return
+	}
+
+	if err := valueRemove(context.Background(), tmpdir, key); err != nil {
+		t.Errorf("error: %+v", err)
+		//return
+	}
+}

engine/metaparams.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,7 +13,19 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package engine
 
@@ -22,8 +34,8 @@ import (
 	"strconv"
 
 	"github.com/purpleidea/mgmt/util"
+	"github.com/purpleidea/mgmt/util/errwrap"
 
-	errwrap "github.com/pkg/errors"
 	"golang.org/x/time/rate"
 )
 
@@ -36,7 +48,10 @@ var DefaultMetaParams = &MetaParams{
 	Poll:  0,        // defaults to watching for events
 	Limit: rate.Inf, // defaults to no limit
 	Burst: 0,        // no burst needed on an infinite rate
+	Reset: false,
 	//Sema:  []string{},
+	Rewatch: false,
+	Realize: false, // true would be more awesome, but unexpected for users
 }
 
 // MetaRes is the interface a resource must implement to support meta params.
@@ -44,6 +59,10 @@ var DefaultMetaParams = &MetaParams{
 type MetaRes interface {
 	// MetaParams lets you get or set meta params for the resource.
 	MetaParams() *MetaParams
+
+	// SetMetaParams lets you set all of the meta params for the resource in
+	// a single call.
+	SetMetaParams(*MetaParams)
 }
 
 // MetaParams provides some meta parameters that apply to every resource.
@@ -58,9 +77,16 @@ type MetaParams struct {
 	// reason to want to do something differently for the Watch errors.
 
 	// Retry is the number of times to retry on error. Use -1 for infinite.
+	// This value is used for both Watch and CheckApply.
 	Retry int16 `yaml:"retry"`
 
-	// Delay is the number of milliseconds to wait between retries.
+	// RetryReset resets the retry count for CheckApply if it succeeds. This
+	// value is currently different from the count used for Watch.
+	// TODO: Consider resetting retry count for watch if it sends an event?
+	RetryReset bool `yaml:"retryreset"`
+
+	// Delay is the number of milliseconds to wait between retries. This
+	// value is used for both Watch and CheckApply.
 	Delay uint64 `yaml:"delay"`
 
 	// Poll is the number of seconds between poll intervals. Use 0 to Watch.
@@ -72,11 +98,40 @@ type MetaParams struct {
 	// Burst is the number of events to allow in a burst.
 	Burst int `yaml:"burst"`
 
+	// Reset causes the meta param state to reset when the resource changes.
+	// What this means is if you have a resource of a specific kind and name
+	// and in the subsequent graph it changes because one of its params
+	// changed, normally the Retry, and other params will remember their
+	// state, and you'll not reset the retry counter, however if this is
+	// true, then it will get reset. Note that any normal reset mechanisms
+	// built into retry are not affected by this.
+	Reset bool `yaml:"reset"`
+
 	// Sema is a list of semaphore ids in the form `id` or `id:count`. If
 	// you don't specify a count, then 1 is assumed. The sema of `foo` which
 	// has a count equal to 1, is different from a sema named `foo:1` which
 	// also has a count equal to 1, but is a different semaphore.
 	Sema []string `yaml:"sema"`
+
+	// Rewatch specifies whether we re-run the Watch worker during a swap
+	// if it has errored. When doing a GraphCmp to swap the graphs, if this
+	// is true, and this particular worker has errored, then we'll remove it
+	// and add it back as a new vertex, thus causing it to run again. This
+	// is different from the Retry metaparam which applies during the normal
+	// execution. It is only when this is exhausted that we're in permanent
+	// worker failure, and only then can we rely on this metaparam. This is
+	// false by default, as the frequency of graph changes makes it unlikely
+	// that you wanted this enabled on most resources.
+	Rewatch bool `yaml:"rewatch"`
+
+	// Realize ensures that the resource is guaranteed to converge at least
+	// once before a potential graph swap removes or changes it. This
+	// guarantee is useful for fast changing graphs, to ensure that the
+	// brief creation of a resource is seen. This guarantee does not prevent
+	// against the engine quitting normally, and it can't guarantee it if
+	// the resource is blocked because of a failed pre-requisite resource.
+	// XXX: Not implemented!
+	Realize bool `yaml:"realize"`
 }
 
 // Cmp compares two AutoGroupMeta structs and determines if they're equivalent.
@@ -109,11 +164,21 @@ func (obj *MetaParams) Cmp(meta *MetaParams) error {
 	if obj.Burst != meta.Burst {
 		return fmt.Errorf("values for Burst are different")
 	}
+	if obj.Reset != meta.Reset {
+		return fmt.Errorf("values for Reset are different")
+	}
 
 	if err := util.SortedStrSliceCompare(obj.Sema, meta.Sema); err != nil {
 		return errwrap.Wrapf(err, "values for Sema are different")
 	}
 
+	if obj.Rewatch != meta.Rewatch {
+		return fmt.Errorf("values for Rewatch are different")
+	}
+	if obj.Realize != meta.Realize {
+		return fmt.Errorf("values for Realize are different")
+	}
+
 	return nil
 }
 
@@ -143,13 +208,16 @@ func (obj *MetaParams) Copy() *MetaParams {
 		copy(sema, obj.Sema)
 	}
 	return &MetaParams{
-		Noop:  obj.Noop,
-		Retry: obj.Retry,
-		Delay: obj.Delay,
-		Poll:  obj.Poll,
-		Limit: obj.Limit, // FIXME: can we copy this type like this? test me!
-		Burst: obj.Burst,
-		Sema:  sema,
+		Noop:    obj.Noop,
+		Retry:   obj.Retry,
+		Delay:   obj.Delay,
+		Poll:    obj.Poll,
+		Limit:   obj.Limit, // FIXME: can we copy this type like this? test me!
+		Burst:   obj.Burst,
+		Reset:   obj.Reset,
+		Sema:    sema,
+		Rewatch: obj.Rewatch,
+		Realize: obj.Realize,
 	}
 }
 
@@ -167,3 +235,18 @@ func (obj *MetaParams) UnmarshalYAML(unmarshal func(interface{}) error) error {
 	*obj = MetaParams(raw) // restore from indirection with type conversion!
 	return nil
 }
+
+// MetaState is some local meta param state that is saved between resources of
+// the same kind+name unique ID. Even though the vertex/resource pointer might
+// change during a graph switch (because the params changed) we might be
+// logically referring to the same resource, and we might want to preserve some
+// data across that switch. The common example is the resource retry count. If a
+// resource failed three times, and we had a limit of five before it would be a
+// permanent failure, then we don't want to reset this counter just because we
+// changed a parameter (field) of the resource. This doesn't mean we don't want
+// to ever reset these counts. For that, flip on the reset meta param.
+type MetaState struct {
+
+	// CheckApplyRetry is the current retry count for CheckApply.
+	CheckApplyRetry int16
+}

engine/metaparams_test.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,9 +13,21 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
-// +build !root
+//go:build !root
 
 package engine
 

engine/refresh.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,7 +13,19 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package engine
 

engine/resources.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,25 +13,39 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package engine
 
 import (
+	"context"
 	"encoding/gob"
 	"fmt"
 
-	"github.com/purpleidea/mgmt/engine/event"
+	"github.com/purpleidea/mgmt/engine/local"
+	"github.com/purpleidea/mgmt/pgraph"
+	"github.com/purpleidea/mgmt/util/errwrap"
 
-	errwrap "github.com/pkg/errors"
 	"gopkg.in/yaml.v2"
 )
 
 // TODO: should each resource be a sub-package?
 var registeredResources = map[string]func() Res{}
 
-// RegisterResource registers a new resource by providing a constructor
-// function that returns a resource object ready to be unmarshalled from YAML.
+// RegisterResource registers a new resource by providing a constructor function
+// that returns a resource object ready to be unmarshalled from YAML.
 func RegisterResource(kind string, fn func() Res) {
 	f := fn()
 	if kind == "" {
@@ -87,28 +101,19 @@ type Init struct {
 	// Program is the name of the program.
 	Program string
 
+	// Version is the version of the program.
+	Version string
+
 	// Hostname is the uuid for the host.
 	Hostname string
 
 	// Called from within Watch:
 
 	// Running must be called after your watches are all started and ready.
-	Running func() error
+	Running func()
 
 	// Event sends an event notifying the engine of a possible state change.
-	Event func() error
-
-	// Events returns a channel that we must watch for messages from the
-	// engine. When it closes, this is a signal to shutdown.
-	Events chan event.Kind
-
-	// Read processes messages that come in from the Events channel. It is a
-	// helper method that knows how to handle the pause mechanism correctly.
-	Read func(event.Kind) error
-
-	// Dirty marks the resource state as dirty. This signals to the engine
-	// that CheckApply will have some work to do in order to converge it.
-	Dirty func()
+	Event func()
 
 	// Called from within CheckApply:
 
@@ -129,6 +134,25 @@ type Init struct {
 
 	// Other functionality:
 
+	// Graph is a function that returns the current graph. The returned
+	// value won't be valid after a graphsync so make sure to call this when
+	// you are about to use it, and discard it right after.
+	// FIXME: it might be better to offer a safer, more limited, GraphQuery?
+	//Graph func() *pgraph.Graph // TODO: not implemented, use FilteredGraph
+
+	// FilteredGraph is a function that returns a filtered variant of the
+	// current graph. Only resource that have allowed themselves to be added
+	// into this graph will appear. If they did not consent, then those
+	// vertices and any associated edges, will not be present.
+	FilteredGraph func() (*pgraph.Graph, error)
+
+	// TODO: GraphQuery offers an interface to query the resource graph.
+
+	// Local has a bunch of methods and properties which are useful for
+	// operations on the local machine and for communication between
+	// functions and resources.
+	Local *local.API
+
 	// World provides a connection to the outside world. This is most often
 	// used for communicating with the distributed database.
 	World World
@@ -184,20 +208,26 @@ type Res interface {
 	// and data from the engine.
 	Init(*Init) error
 
-	// Close is run by the engine to clean up after the resource is done.
-	Close() error
+	// Cleanup is run by the engine to clean up after the resource is done.
+	Cleanup() error
 
 	// Watch is run by the engine to monitor for state changes. If it
 	// detects any, it notifies the engine which will usually run CheckApply
-	// in response.
-	Watch() error
+	// in response. If the input context cancels, we must shutdown.
+	Watch(context.Context) error
 
-	// CheckApply determines if the state of the resource is connect and if
-	// asked to with the `apply` variable, applies the requested state.
-	CheckApply(apply bool) (checkOK bool, err error)
+	// CheckApply determines if the state of the resource is correct and if
+	// asked to with the `apply` variable, applies the requested state. If
+	// the input context cancels, we must return as quickly as possible. We
+	// should never exit immediately if this would cause permanent
+	// corruption of some sort. However it doesn't mean that a resource was
+	// taken to the desired state.
+	CheckApply(ctx context.Context, apply bool) (checkOK bool, err error)
 
 	// Cmp compares itself to another resource and returns an error if they
-	// are not equivalent.
+	// are not equivalent. This is more strict than the Adapts method of the
+	// CompatibleRes interface which allows for equivalent differences if
+	// the have a compatible result in CheckApply.
 	Cmp(Res) error
 }
 
@@ -212,6 +242,22 @@ func Stringer(res Res) string {
 	return Repr(res.Kind(), res.Name())
 }
 
+// ResPtrUID is a unique identifier that is consistent for the kind and name of
+// the resource only. This was formerly a string, but a struct is more precise.
+// The result is suitable as a unique map key.
+type ResPtrUID struct {
+	kind string
+	name string
+}
+
+// PtrUID generates a ResPtrUID from a resource. The result is suitable as a
+// unique map key.
+func PtrUID(res Res) ResPtrUID {
+	// the use of "repr" is kind of arbitrary as long as it's unique
+	//return ResPtrUID(Repr(res.Kind(), res.Name()))
+	return ResPtrUID{kind: res.Kind(), name: res.Name()}
+}
+
 // Validate validates a resource by checking multiple aspects. This is the main
 // entry point for running all the validation steps on a resource.
 func Validate(res Res) error {
@@ -234,8 +280,8 @@ func Validate(res Res) error {
 // the Interrupt method to shutdown the resource quickly. Running this method
 // may leave the resource in a partial state, however this may be desired if you
 // want a faster exit or if you'd prefer a partial state over letting the
-// resource complete in a situation where you made an error and you wish to
-// exit quickly to avoid data loss. It is usually triggered after multiple ^C
+// resource complete in a situation where you made an error and you wish to exit
+// quickly to avoid data loss. It is usually triggered after multiple ^C
 // signals.
 type InterruptableRes interface {
 	Res
@@ -246,15 +292,50 @@ type InterruptableRes interface {
 	// is designed to unblock any long running operation that is occurring
 	// in the CheckApply portion of the life cycle. If the resource has
 	// already exited, running this method should not block. (That is to say
-	// that you should not expect CheckApply or Watch to be able to alive
-	// and able to read from a channel to satisfy your request.) It is best
-	// to probably have this close a channel to multicast that signal around
-	// to anyone who can detect it in a select. If you are in a situation
-	// which cannot interrupt, then you can return an error.
+	// that you should not expect CheckApply or Watch to be alive and be
+	// able to read from a channel to satisfy your request.) It is best to
+	// probably have this close a channel to multicast that signal around to
+	// anyone who can detect it in a select. If you are in a situation which
+	// cannot interrupt, then you can return an error.
 	// FIXME: implement, and check the above description is what we expect!
 	Interrupt() error
 }
 
+// CopyableRes is an interface that a resource can implement if we want to be
+// able to copy the resource to build another one.
+type CopyableRes interface {
+	Res
+
+	// Copy returns a new resource which has a copy of the public data.
+	// Don't call this directly, use engine.ResCopy instead.
+	// TODO: should we copy any private state or not?
+	Copy() CopyableRes
+}
+
+// CompatibleRes is an interface that a resource can implement to express if a
+// similar variant of itself is functionally equivalent. For example, two `pkg`
+// resources that install `cowsay` could be equivalent if one requests a state
+// of `installed` and the other requests `newest`, since they'll finish with a
+// compatible result. This doesn't need to be behind a metaparam flag or trait,
+// because it is never beneficial to turn it off, unless there is a bug to fix.
+type CompatibleRes interface {
+	//Res // causes "duplicate method" error
+	CopyableRes // we'll need to use the Copy method in the Merge function!
+
+	// Adapts compares itself to another resource and returns an error if
+	// they are not compatibly equivalent. This is less strict than the
+	// default `Cmp` method which should be used for most cases. Don't call
+	// this directly, use engine.AdaptCmp instead.
+	Adapts(CompatibleRes) error
+
+	// Merge returns the combined resource to use when two are equivalent.
+	// This might get called multiple times for N different resources that
+	// need to get merged, and so it should produce a consistent result no
+	// matter which order it is called in. Don't call this directly, use
+	// engine.ResMerge instead.
+	Merge(CompatibleRes) (CompatibleRes, error)
+}
+
 // CollectableRes is an interface for resources that support collection. It is
 // currently temporary until a proper API for all resources is invented.
 type CollectableRes interface {

engine/resources/augeas.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,24 +13,35 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
-// +build !noaugeas
+//go:build !noaugeas
 
 package resources
 
 import (
+	"context"
 	"fmt"
 	"os"
 	"strings"
 
 	"github.com/purpleidea/mgmt/engine"
 	"github.com/purpleidea/mgmt/engine/traits"
-	"github.com/purpleidea/mgmt/recwatch"
+	"github.com/purpleidea/mgmt/util/errwrap"
+	"github.com/purpleidea/mgmt/util/recwatch"
 
-	errwrap "github.com/pkg/errors"
-	// FIXME: we vendor go/augeas because master requires augeas 1.6.0
-	// and libaugeas-dev-1.6.0 is not yet available in a PPA.
 	"honnef.co/go/augeas"
 )
 
@@ -51,24 +62,24 @@ type AugeasRes struct {
 	init *engine.Init
 
 	// File is the path to the file targeted by this resource.
-	File string `yaml:"file"`
+	File string `lang:"file" yaml:"file"`
 
 	// Lens is the lens used by this resource. If specified, mgmt
 	// will lower the augeas overhead by only loading that lens.
-	Lens string `yaml:"lens"`
+	Lens string `lang:"lens" yaml:"lens"`
 
-	// Sets is a list of changes that will be applied to the file, in the form of
-	// ["path", "value"]. mgmt will run augeas.Get() before augeas.Set(), to
-	// prevent changing the file when it is not needed.
-	Sets []*AugeasSet `yaml:"sets"`
+	// Sets is a list of changes that will be applied to the file, in the
+	// form of ["path", "value"]. mgmt will run augeas.Get() before
+	// augeas.Set(), to prevent changing the file when it is not needed.
+	Sets []*AugeasSet `lang:"sets" yaml:"sets"`
 
 	recWatcher *recwatch.RecWatcher // used to watch the changed files
 }
 
 // AugeasSet represents a key/value pair of settings to be applied.
 type AugeasSet struct {
-	Path  string `yaml:"path"`  // The relative path to the value to be changed.
-	Value string `yaml:"value"` // The value to be set on the given Path.
+	Path  string `lang:"path" yaml:"path"`   // The relative path to the value to be changed.
+	Value string `lang:"value" yaml:"value"` // The value to be set on the given Path.
 }
 
 // Cmp compares this set with another one.
@@ -119,15 +130,15 @@ func (obj *AugeasRes) Init(init *engine.Init) error {
 	return nil
 }
 
-// Close is run by the engine to clean up after the resource is done.
-func (obj *AugeasRes) Close() error {
+// Cleanup is run by the engine to clean up after the resource is done.
+func (obj *AugeasRes) Cleanup() error {
 	return nil
 }
 
-// Watch is the primary listener for this resource and it outputs events.
-// Taken from the File resource.
+// Watch is the primary listener for this resource and it outputs events. This
+// was taken from the File resource.
 // FIXME: DRY - This is taken from the file resource
-func (obj *AugeasRes) Watch() error {
+func (obj *AugeasRes) Watch(ctx context.Context) error {
 	var err error
 	obj.recWatcher, err = recwatch.NewRecWatcher(obj.File, false)
 	if err != nil {
@@ -135,10 +146,7 @@ func (obj *AugeasRes) Watch() error {
 	}
 	defer obj.recWatcher.Close()
 
-	// notify engine that we're running
-	if err := obj.init.Running(); err != nil {
-		return err // exit if requested
-	}
+	obj.init.Running() // when started, notify engine that we're running
 
 	var send = false // send event?
 	for {
@@ -158,29 +166,21 @@ func (obj *AugeasRes) Watch() error {
 				obj.init.Logf("Event(%s): %v", event.Body.Name, event.Body.Op)
 			}
 			send = true
-			obj.init.Dirty() // dirty
 
-		case event, ok := <-obj.init.Events:
-			if !ok {
-				return nil
-			}
-			if err := obj.init.Read(event); err != nil {
-				return err
-			}
+		case <-ctx.Done(): // closed by the engine to signal shutdown
+			return nil
 		}
 
 		// do all our event sending all together to avoid duplicate msgs
 		if send {
 			send = false
-			if err := obj.init.Event(); err != nil {
-				return err // exit if requested
-			}
+			obj.init.Event() // notify engine of an event (this can block)
 		}
 	}
 }
 
 // checkApplySet runs CheckApply for one element of the AugeasRes.Set
-func (obj *AugeasRes) checkApplySet(apply bool, ag *augeas.Augeas, set *AugeasSet) (bool, error) {
+func (obj *AugeasRes) checkApplySet(ctx context.Context, apply bool, ag *augeas.Augeas, set *AugeasSet) (bool, error) {
 	fullpath := fmt.Sprintf("/files/%v/%v", obj.File, set.Path)
 
 	// We do not check for errors because errors are also thrown when
@@ -205,7 +205,7 @@ func (obj *AugeasRes) checkApplySet(apply bool, ag *augeas.Augeas, set *AugeasSe
 }
 
 // CheckApply method for Augeas resource.
-func (obj *AugeasRes) CheckApply(apply bool) (bool, error) {
+func (obj *AugeasRes) CheckApply(ctx context.Context, apply bool) (bool, error) {
 	obj.init.Logf("CheckApply: %s", obj.File)
 	// By default we do not set any option to augeas, we use the defaults.
 	opts := augeas.None
@@ -242,7 +242,7 @@ func (obj *AugeasRes) CheckApply(apply bool) (bool, error) {
 
 	checkOK := true
 	for _, set := range obj.Sets {
-		if setCheckOK, err := obj.checkApplySet(apply, &ag, set); err != nil {
+		if setCheckOK, err := obj.checkApplySet(ctx, apply, &ag, set); err != nil {
 			return false, errwrap.Wrapf(err, "augeas: error during CheckApply of one Set")
 		} else if !setCheckOK {
 			checkOK = false
@@ -312,8 +312,8 @@ func (obj *AugeasRes) UIDs() []engine.ResUID {
 	return []engine.ResUID{x}
 }
 
-// UnmarshalYAML is the custom unmarshal handler for this struct.
-// It is primarily useful for setting the defaults.
+// UnmarshalYAML is the custom unmarshal handler for this struct. It is
+// primarily useful for setting the defaults.
 func (obj *AugeasRes) UnmarshalYAML(unmarshal func(interface{}) error) error {
 	type rawRes AugeasRes // indirection to avoid infinite recursion
 

engine/resources/aws_ec2.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2018+ James Shubin and the project contributors
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -13,7 +13,19 @@
 // GNU General Public License for more details.
 //
 // You should have received a copy of the GNU General Public License
-// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
 
 package resources
 
@@ -24,7 +36,7 @@ import (
 	"encoding/json"
 	"encoding/pem"
 	"fmt"
-	"io/ioutil"
+	"io"
 	"net"
 	"net/http"
 	"regexp"
@@ -34,6 +46,7 @@ import (
 
 	"github.com/purpleidea/mgmt/engine"
 	"github.com/purpleidea/mgmt/engine/traits"
+	"github.com/purpleidea/mgmt/util/errwrap"
 
 	"github.com/aws/aws-sdk-go/aws"
 	"github.com/aws/aws-sdk-go/aws/awserr"
@@ -42,8 +55,6 @@ import (
 	cwe "github.com/aws/aws-sdk-go/service/cloudwatchevents"
 	"github.com/aws/aws-sdk-go/service/ec2"
 	"github.com/aws/aws-sdk-go/service/sns"
-	multierr "github.com/hashicorp/go-multierror"
-	errwrap "github.com/pkg/errors"
 )
 
 func init() {
@@ -122,8 +133,8 @@ const (
 )
 
 // AwsRegions is a list of all AWS regions generated using ec2.DescribeRegions.
-// cn-north-1 and us-gov-west-1 are not returned, probably due to security.
-// List available at http://docs.aws.amazon.com/general/latest/gr/rande.html
+// cn-north-1 and us-gov-west-1 are not returned, probably due to security. List
+// available at http://docs.aws.amazon.com/general/latest/gr/rande.html
 var AwsRegions = []string{
 	"ap-northeast-1",
 	"ap-northeast-2",
@@ -148,26 +159,42 @@ var AwsRegions = []string{
 // http://docs.aws.amazon.com/cli/latest/userguide/cli-config-files.html
 type AwsEc2Res struct {
 	traits.Base // add the base methods without re-implementation
+	traits.Sendable
 
 	init *engine.Init
 
-	State   string `yaml:"state"`   // state: running, stopped, terminated
-	Region  string `yaml:"region"`  // region must match an element of AwsRegions
-	Type    string `yaml:"type"`    // type of ec2 instance, eg: t2.micro
-	ImageID string `yaml:"imageid"` // imageid must be available on the chosen region
+	// State must be running, stopped, or terminated.
+	State string `lang:"state" yaml:"state"`
+
+	// Region must match one of the AwsRegions. This list is static at the
+	// moment.
+	Region string `lang:"region" yaml:"region"`
+
+	// Type of ec2 instance, eg: t2.micro for example.
+	Type string `lang:"type" yaml:"type"`
+
+	// ImageID to use, and note that it must be available on the chosen
+	// region.
+	ImageID string `lang:"imageid" yaml:"imageid"`
+
+	// WatchEndpoint is the public url of the sns endpoint, eg:
+	// http://server:12345/ for example.
+	WatchEndpoint string `lang:"watchendpoint" yaml:"watchendpoint"`
+
+	// WatchListenAddr is the local address or port that the sns listens on,
+	// eg: 10.0.0.0:23456 or 23456.
+	WatchListenAddr string `lang:"watchlistenaddr" yaml:"watchlistenaddr"`
 
-	WatchEndpoint   string `yaml:"watchendpoint"`   // the public url of the sns endpoint, eg: http://server:12345/
-	WatchListenAddr string `yaml:"watchlistenaddr"` // the local address or port that the sns listens on, eg: 10.0.0.0:23456 or 23456
 	// ErrorOnMalformedPost controls whether or not malformed HTTP post
 	// requests, that cause JSON decoder errors, will also make the engine
 	// shut down. If ErrorOnMalformedPost set to true and an error occurs,
 	// Watch() will return the error and the engine will shut down.
-	ErrorOnMalformedPost bool `yaml:"erroronmalformedpost"`
+	ErrorOnMalformedPost bool `lang:"erroronmalformedpost" yaml:"erroronmalformedpost"`
 
 	// UserData is used to run bash and cloud-init commands on first launch.
 	// See http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/user-data.html
 	// for documantation and examples.
-	UserData string `yaml:"userdata"`
+	UserData string `lang:"userdata" yaml:"userdata"`
 
 	client *ec2.EC2 // client session for AWS API calls
 
@@ -188,7 +215,8 @@ type AwsEc2Res struct {
 	InstanceID  string
 }
 
-// chanStruct defines the type for a channel used to pass events and errors to watch.
+// chanStruct defines the type for a channel used to pass events and errors to
+// watch.
 type chanStruct struct {
 	event awsEc2Event
 	state string
@@ -234,7 +262,8 @@ type ruleDetail struct {
 	State []string `json:"state"`
 }
 
-// postData is the format of the messages received and decoded by snsPostHandler().
+// postData is the format of the messages received and decoded by
+// snsPostHandler().
 type postData struct {
 	Type             string `json:"Type"`
 	MessageID        string `json:"MessageId"`
@@ -248,7 +277,8 @@ type postData struct {
 	SigningCertURL   string `json:"SigningCertURL"`
 }
 
-// postMsg is used to unmarshal the postData message if it's an event notification.
+// postMsg is used to unmarshal the postData message if it's an event
+// notification.
 type postMsg struct {
 	InstanceID string `json:"instance-id"`
 	State      string `json:"state"`
@@ -386,49 +416,46 @@ func (obj *AwsEc2Res) Init(init *engine.Init) error {
 	return nil
 }
 
-// Close cleans up when we're done. This is needed to delete some of the AWS
+// Cleanup cleans up when we're done. This is needed to delete some of the AWS
 // objects created for the SNS endpoint.
-func (obj *AwsEc2Res) Close() error {
+func (obj *AwsEc2Res) Cleanup() error {
 	var errList error
+	// XXX: do these in a defer of Watch instead.
 	// clean up sns objects created by Init/snsWatch
 	if obj.snsClient != nil {
 		// delete the topic and associated subscriptions
-		if err := obj.snsDeleteTopic(obj.snsTopicArn); err != nil {
-			errList = multierr.Append(errList, err)
-		}
+		e1 := obj.snsDeleteTopic(obj.snsTopicArn)
+		errList = errwrap.Append(errList, e1)
 		// remove the target
-		if err := obj.cweRemoveTarget(CweTargetID, CweRuleName); err != nil {
-			errList = multierr.Append(errList, err)
-		}
+		e2 := obj.cweRemoveTarget(CweTargetID, CweRuleName)
+		errList = errwrap.Append(errList, e2)
 		// delete the cloudwatch rule
-		if err := obj.cweDeleteRule(CweRuleName); err != nil {
-			errList = multierr.Append(errList, err)
-		}
+		e3 := obj.cweDeleteRule(CweRuleName)
+		errList = errwrap.Append(errList, e3)
 	}
 
 	return errList
 }
 
 // Watch is the primary listener for this resource and it outputs events.
-func (obj *AwsEc2Res) Watch() error {
+func (obj *AwsEc2Res) Watch(ctx context.Context) error {
 	if obj.WatchListenAddr != "" {
-		return obj.snsWatch()
+		return obj.snsWatch(ctx)
 	}
-	return obj.longpollWatch()
+	return obj.longpollWatch(ctx)
 }
 
-// longpollWatch uses the ec2 api's built in methods to watch ec2 resource state.
-func (obj *AwsEc2Res) longpollWatch() error {
+// longpollWatch uses the ec2 api's built in methods to watch ec2 resource
+// state.
+func (obj *AwsEc2Res) longpollWatch(ctx context.Context) error {
 	send := false
 
 	// We tell the engine that we're running right away. This is not correct,
 	// but the api doesn't have a way to signal when the waiters are ready.
-	if err := obj.init.Running(); err != nil {
-		return err // exit if requested
-	}
+	obj.init.Running() // when started, notify engine that we're running
 
 	// cancellable context used for exiting cleanly
-	ctx, cancel := context.WithCancel(context.TODO())
+	innerCtx, cancel := context.WithCancel(context.TODO())
 
 	// clean up when we're done
 	defer obj.wg.Wait()
@@ -463,7 +490,7 @@ func (obj *AwsEc2Res) longpollWatch() error {
 			}
 
 			// wait for the instance state to change
-			state, err := stateWaiter(ctx, instance, obj.client)
+			state, err := stateWaiter(innerCtx, instance, obj.client)
 			if err != nil {
 				select {
 				case obj.awsChan <- &chanStruct{
@@ -488,14 +515,6 @@ func (obj *AwsEc2Res) longpollWatch() error {
 	// process events from the goroutine
 	for {
 		select {
-		case event, ok := <-obj.init.Events:
-			if !ok {
-				return nil
-			}
-			if err := obj.init.Read(event); err != nil {
-				return err
-			}
-
 		case msg, ok := <-obj.awsChan:
 			if !ok {
 				return nil
@@ -509,25 +528,26 @@ func (obj *AwsEc2Res) longpollWatch() error {
 				continue
 			default:
 				obj.init.Logf("State: %v", msg.state)
-				obj.init.Dirty() // dirty
 				send = true
 			}
+
+		case <-ctx.Done(): // closed by the engine to signal shutdown
+			return nil
 		}
+
 		if send {
 			send = false
-			if err := obj.init.Event(); err != nil {
-				return err // exit if requested
-			}
+			obj.init.Event() // notify engine of an event (this can block)
 		}
 	}
 }
 
 // snsWatch uses amazon's SNS and CloudWatchEvents APIs to get instance state-
-// change notifications pushed to the http endpoint (snsServer) set up below.
-// In Init() a CloudWatch rule is created along with a corresponding SNS topic
-// that it can publish to. snsWatch creates an http server which listens for
-// messages published to the topic and processes them accordingly.
-func (obj *AwsEc2Res) snsWatch() error {
+// change notifications pushed to the http endpoint (snsServer) set up below. In
+// Init() a CloudWatch rule is created along with a corresponding SNS topic that
+// it can publish to. snsWatch creates an http server which listens for messages
+// published to the topic and processes them accordingly.
+func (obj *AwsEc2Res) snsWatch(ctx context.Context) error {
 	send := false
 	defer obj.wg.Wait()
 	// create the sns listener
@@ -542,9 +562,9 @@ func (obj *AwsEc2Res) snsWatch() error {
 	}
 	// close the listener and shutdown the sns server when we're done
 	defer func() {
-		ctx, cancel := context.WithTimeout(context.TODO(), SnsServerShutdownTimeout*time.Second)
+		innerCtx, cancel := context.WithTimeout(context.TODO(), SnsServerShutdownTimeout*time.Second)
 		defer cancel()
-		if err := snsServer.Shutdown(ctx); err != nil {
+		if err := snsServer.Shutdown(innerCtx); err != nil {
 			if err != context.Canceled {
 				obj.init.Logf("error stopping sns endpoint: %s", err)
 				return
@@ -587,14 +607,6 @@ func (obj *AwsEc2Res) snsWatch() error {
 	// process events
 	for {
 		select {
-		case event, ok := <-obj.init.Events:
-			if !ok {
-				return nil
-			}
-			if err := obj.init.Read(event); err != nil {
-				return err
-			}
-
 		case msg, ok := <-obj.awsChan:
 			if !ok {
 				return nil
@@ -607,26 +619,25 @@ func (obj *AwsEc2Res) snsWatch() error {
 			// is confirmed, we are ready to receive events, so we
 			// can notify the engine that we're running.
 			if msg.event == awsEc2EventWatchReady {
-				if err := obj.init.Running(); err != nil {
-					return err // exit if requested
-				}
+				obj.init.Running() // when started, notify engine that we're running
 				continue
 			}
 			obj.init.Logf("State: %v", msg.event)
-			obj.init.Dirty() // dirty
 			send = true
+
+		case <-ctx.Done(): // closed by the engine to signal shutdown
+			return nil
 		}
+
 		if send {
 			send = false
-			if err := obj.init.Event(); err != nil {
-				return err // exit if requested
-			}
+			obj.init.Event() // notify engine of an event (this can block)
 		}
 	}
 }
 
 // CheckApply method for AwsEc2 resource.
-func (obj *AwsEc2Res) CheckApply(apply bool) (checkOK bool, err error) {
+func (obj *AwsEc2Res) CheckApply(ctx context.Context, apply bool) (bool, error) {
 	obj.init.Logf("CheckApply(%t)", apply)
 
 	// find the instance we need to check
@@ -736,7 +747,7 @@ func (obj *AwsEc2Res) CheckApply(apply bool) (checkOK bool, err error) {
 	}
 
 	// context to cancel the waiter if it takes too long
-	ctx, cancel := context.WithTimeout(context.TODO(), waitTimeout*time.Second)
+	innerCtx, cancel := context.WithTimeout(ctx, waitTimeout*time.Second)
 	defer cancel()
 
 	// wait until the state converges
@@ -745,11 +756,11 @@ func (obj *AwsEc2Res) CheckApply(apply bool) (checkOK bool, err error) {
 	}
 	switch obj.State {
 	case ec2.InstanceStateNameRunning:
-		err = obj.client.WaitUntilInstanceRunningWithContext(ctx, waitInput)
+		err = obj.client.WaitUntilInstanceRunningWithContext(innerCtx, waitInput)
 	case ec2.InstanceStateNameStopped:
-		err = obj.client.WaitUntilInstanceStoppedWithContext(ctx, waitInput)
+		err = obj.client.WaitUntilInstanceStoppedWithContext(innerCtx, waitInput)
 	case ec2.InstanceStateNameTerminated:
-		err = obj.client.WaitUntilInstanceTerminatedWithContext(ctx, waitInput)
+		err = obj.client.WaitUntilInstanceTerminatedWithContext(innerCtx, waitInput)
 	default:
 		return false, errwrap.Wrapf(err, "unrecognized instance state")
 	}
@@ -773,45 +784,37 @@ func (obj *AwsEc2Res) CheckApply(apply bool) (checkOK bool, err error) {
 
 // Cmp compares two resources and returns an error if they are not equivalent.
 func (obj *AwsEc2Res) Cmp(r engine.Res) error {
-	if !obj.Compare(r) {
-		return fmt.Errorf("did not compare")
-	}
-	return nil
-}
-
-// Compare two resources and return if they are equivalent.
-func (obj *AwsEc2Res) Compare(r engine.Res) bool {
 	// we can only compare AwsEc2Res to others of the same resource kind
 	res, ok := r.(*AwsEc2Res)
 	if !ok {
-		return false
+		return fmt.Errorf("not a %s", obj.Kind())
 	}
 
 	if obj.State != res.State {
-		return false
+		return fmt.Errorf("the State differs")
 	}
 	if obj.Region != res.Region {
-		return false
+		return fmt.Errorf("the Region differs")
 	}
 	if obj.Type != res.Type {
-		return false
+		return fmt.Errorf("the Type differs")
 	}
 	if obj.ImageID != res.ImageID {
-		return false
+		return fmt.Errorf("the ImageID differs")
 	}
 	if obj.WatchEndpoint != res.WatchEndpoint {
-		return false
+		return fmt.Errorf("the WatchEndpoint differs")
 	}
 	if obj.WatchListenAddr != res.WatchListenAddr {
-		return false
+		return fmt.Errorf("the WatchListenAddr differs")
 	}
 	if obj.ErrorOnMalformedPost != res.ErrorOnMalformedPost {
-		return false
+		return fmt.Errorf("the ErrorOnMalformedPost differs")
 	}
 	if obj.UserData != res.UserData {
-		return false
+		return fmt.Errorf("the UserData differs")
 	}
-	return true
+	return nil
 }
 
 func (obj *AwsEc2Res) prependName() string {
@@ -825,8 +828,8 @@ type AwsEc2UID struct {
 	name string
 }
 
-// UIDs includes all params to make a unique identification of this object.
-// Most resources only return one, although some resources can return multiple.
+// UIDs includes all params to make a unique identification of this object. Most
+// resources only return one, although some resources can return multiple.
 func (obj *AwsEc2Res) UIDs() []engine.ResUID {
 	x := &AwsEc2UID{
 		BaseUID: engine.BaseUID{Name: obj.Name(), Kind: obj.Kind()},
@@ -835,8 +838,8 @@ func (obj *AwsEc2Res) UIDs() []engine.ResUID {
 	return []engine.ResUID{x}
 }
 
-// UnmarshalYAML is the custom unmarshal handler for this struct.
-// It is primarily useful for setting the defaults.
+// UnmarshalYAML is the custom unmarshal handler for this struct. It is
+// primarily useful for setting the defaults.
 func (obj *AwsEc2Res) UnmarshalYAML(unmarshal func(interface{}) error) error {
 	type rawRes AwsEc2Res // indirection to avoid infinite recursion
 
@@ -972,8 +975,8 @@ func (obj *AwsEc2Res) snsVerifySignature(post postData) error {
 	return nil
 }
 
-// snsGetCert downloads and parses the signing certificate from the provided
-// URL for message verification.
+// snsGetCert downloads and parses the signing certificate from the provided URL
+// for message verification.
 func (obj *AwsEc2Res) snsGetCert(url string) (*x509.Certificate, error) {
 	// only download valid certificates from amazon
 	matchURL, err := regexp.MatchString(SnsCertURLRegex, url)
@@ -989,7 +992,7 @@ func (obj *AwsEc2Res) snsGetCert(url string) (*x509.Certificate, error) {
 		return nil, errwrap.Wrapf(err, "http get error")
 	}
 	defer resp.Body.Close()
-	body, err := ioutil.ReadAll(resp.Body)
+	body, err := io.ReadAll(resp.Body)
 	if err != nil {
 		return nil, errwrap.Wrapf(err, "error reading post body")
 	}
@@ -1047,7 +1050,7 @@ func (obj *AwsEc2Res) snsMakeTopic() (string, error) {
 	}
 	obj.init.Logf("Created SNS Topic")
 	if topic.TopicArn == nil {
-		return "", fmt.Errorf("TopicArn is nil")
+		return "", fmt.Errorf("the TopicArn is nil")
 	}
 	return *topic.TopicArn, nil
 }
@@ -1065,8 +1068,8 @@ func (obj *AwsEc2Res) snsDeleteTopic(topicArn string) error {
 	return nil
 }
 
-// snsSubscribe subscribes the endpoint to the sns topic.
-// Returning SubscriptionArn here is useless as it is still pending confirmation.
+// snsSubscribe subscribes the endpoint to the sns topic. Returning
+// SubscriptionArn here is useless as it is still pending confirmation.
 func (obj *AwsEc2Res) snsSubscribe(endpoint string, topicArn string) error {
 	// subscribe to the topic
 	subInput := &sns.SubscribeInput{
@@ -1082,8 +1085,8 @@ func (obj *AwsEc2Res) snsSubscribe(endpoint string, topicArn string) error {
 	return nil
 }
 
-// snsConfirmSubscription confirms the sns subscription.
-// Returning SubscriptionArn here is useless as it is still pending confirmation.
+// snsConfirmSubscription confirms the sns subscription. Returning
+// SubscriptionArn here is useless as it is still pending confirmation.
 func (obj *AwsEc2Res) snsConfirmSubscription(topicArn string, token string) error {
 	// confirm the subscription
 	csInput := &sns.ConfirmSubscriptionInput{
@@ -1135,7 +1138,8 @@ func (obj *AwsEc2Res) snsProcessEvent(message, instanceName string) (awsEc2Event
 	return awsEc2EventNone, nil
 }
 
-// snsAuthorize adds the necessary permission for cloudwatch to publish to the SNS topic.
+// snsAuthorize adds the necessary permission for cloudwatch to publish to the
+// SNS topic.
 func (obj *AwsEc2Res) snsAuthorizeCloudWatch(topicArn string) error {
 	// get the topic attributes, including the security policy
 	gaInput := &sns.GetTopicAttributesInput{

engine/resources/config_etcd.go

@@ -0,0 +1,258 @@
+// Mgmt
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
+
+package resources
+
+import (
+	"context"
+	"fmt"
+	"sync"
+	"time"
+
+	"github.com/purpleidea/mgmt/engine"
+	"github.com/purpleidea/mgmt/engine/traits"
+	"github.com/purpleidea/mgmt/util"
+	"github.com/purpleidea/mgmt/util/errwrap"
+)
+
+func init() {
+	engine.RegisterResource("config:etcd", func() engine.Res { return &ConfigEtcdRes{} })
+}
+
+const (
+	sizeCheckApplyTimeout = 5 * time.Second
+)
+
+// ConfigEtcdRes is a resource that sets mgmt's etcd configuration.
+type ConfigEtcdRes struct {
+	traits.Base // add the base methods without re-implementation
+
+	init *engine.Init
+
+	// IdealClusterSize is the requested minimum size of the cluster. If you
+	// set this to zero, it will cause a cluster wide shutdown if
+	// AllowSizeShutdown is true. If it's not true, then it will cause a
+	// validation error.
+	IdealClusterSize uint16 `lang:"idealclustersize"`
+	// AllowSizeShutdown is a required safety flag that you must set to true
+	// if you want to allow causing a cluster shutdown by setting
+	// IdealClusterSize to zero.
+	AllowSizeShutdown bool `lang:"allow_size_shutdown"`
+
+	// sizeFlag determines whether sizeCheckApply already ran or not.
+	sizeFlag bool
+
+	interruptChan chan struct{}
+}
+
+// Default returns some sensible defaults for this resource.
+func (obj *ConfigEtcdRes) Default() engine.Res {
+	return &ConfigEtcdRes{}
+}
+
+// Validate if the params passed in are valid data.
+func (obj *ConfigEtcdRes) Validate() error {
+	if obj.IdealClusterSize < 0 {
+		return fmt.Errorf("the IdealClusterSize param must be positive")
+	}
+
+	if obj.IdealClusterSize == 0 && !obj.AllowSizeShutdown {
+		return fmt.Errorf("the IdealClusterSize can't be zero if AllowSizeShutdown is false")
+	}
+
+	return nil
+}
+
+// Init runs some startup code for this resource.
+func (obj *ConfigEtcdRes) Init(init *engine.Init) error {
+	obj.init = init // save for later
+
+	obj.interruptChan = make(chan struct{})
+
+	return nil
+}
+
+// Cleanup is run by the engine to clean up after the resource is done.
+func (obj *ConfigEtcdRes) Cleanup() error {
+	return nil
+}
+
+// Watch is the primary listener for this resource and it outputs events.
+func (obj *ConfigEtcdRes) Watch(ctx context.Context) error {
+	wg := &sync.WaitGroup{}
+	defer wg.Wait()
+	innerCtx, cancel := context.WithCancel(ctx)
+	defer cancel()
+	ch, err := obj.init.World.IdealClusterSizeWatch(util.CtxWithWg(innerCtx, wg))
+	if err != nil {
+		return errwrap.Wrapf(err, "could not watch ideal cluster size")
+	}
+
+	obj.init.Running() // when started, notify engine that we're running
+
+Loop:
+	for {
+		select {
+		case event, ok := <-ch:
+			if !ok {
+				break Loop
+			}
+			if obj.init.Debug {
+				obj.init.Logf("event: %+v", event)
+			}
+			// pass through and send an event
+
+		case <-ctx.Done(): // closed by the engine to signal shutdown
+			return nil
+		}
+
+		obj.init.Event() // notify engine of an event (this can block)
+	}
+
+	return nil
+}
+
+// sizeCheckApply sets the IdealClusterSize parameter. If it sees a value change
+// to zero, then it *won't* try and change it away from zero, because it assumes
+// that someone has requested a shutdown. If the value is seen on first startup,
+// then it will change it, because it might be a zero from the previous cluster.
+func (obj *ConfigEtcdRes) sizeCheckApply(ctx context.Context, apply bool) (bool, error) {
+	wg := &sync.WaitGroup{}
+	defer wg.Wait() // this must be above the defer cancel() call
+	ctx, cancel := context.WithTimeout(ctx, sizeCheckApplyTimeout)
+	defer cancel()
+	wg.Add(1)
+	go func() {
+		defer wg.Done()
+		select {
+		case <-obj.interruptChan:
+			cancel()
+		case <-ctx.Done():
+			// let this exit
+		}
+	}()
+
+	val, err := obj.init.World.IdealClusterSizeGet(ctx)
+	if err != nil {
+		return false, errwrap.Wrapf(err, "could not get ideal cluster size")
+	}
+
+	// if we got a value of zero, and we've already run before, then it's ok
+	if obj.IdealClusterSize != 0 && val == 0 && obj.sizeFlag {
+		obj.init.Logf("impending cluster shutdown, not setting ideal cluster size")
+		return true, nil // impending shutdown, don't try and cancel it.
+	}
+	obj.sizeFlag = true
+
+	// must be done after setting the above flag
+	if obj.IdealClusterSize == val { // state is correct
+		return true, nil
+	}
+
+	if !apply {
+		return false, nil
+	}
+
+	// set!
+	// This is run as a transaction so we detect if we needed to change it.
+	changed, err := obj.init.World.IdealClusterSizeSet(ctx, obj.IdealClusterSize)
+	if err != nil {
+		return false, errwrap.Wrapf(err, "could not set ideal cluster size")
+	}
+	if !changed {
+		return true, nil // we lost a race, which means no change needed
+	}
+	obj.init.Logf("set dynamic cluster size to: %d", obj.IdealClusterSize)
+
+	return false, nil
+}
+
+// CheckApply method for Noop resource. Does nothing, returns happy!
+func (obj *ConfigEtcdRes) CheckApply(ctx context.Context, apply bool) (bool, error) {
+	checkOK := true
+
+	if c, err := obj.sizeCheckApply(ctx, apply); err != nil {
+		return false, err
+	} else if !c {
+		checkOK = false
+	}
+
+	// TODO: add more config settings management here...
+	//if c, err := obj.TODOCheckApply(ctx, apply); err != nil {
+	//	return false, err
+	//} else if !c {
+	//	checkOK = false
+	//}
+
+	return checkOK, nil // w00t
+}
+
+// Cmp compares two resources and returns an error if they are not equivalent.
+func (obj *ConfigEtcdRes) Cmp(r engine.Res) error {
+	// we can only compare ConfigEtcdRes to others of the same resource kind
+	res, ok := r.(*ConfigEtcdRes)
+	if !ok {
+		return fmt.Errorf("not a %s", obj.Kind())
+	}
+
+	if obj.IdealClusterSize != res.IdealClusterSize {
+		return fmt.Errorf("the IdealClusterSize param differs")
+	}
+	if obj.AllowSizeShutdown != res.AllowSizeShutdown {
+		return fmt.Errorf("the AllowSizeShutdown param differs")
+	}
+
+	return nil
+}
+
+// Interrupt is called to ask the execution of this resource to end early.
+func (obj *ConfigEtcdRes) Interrupt() error {
+	close(obj.interruptChan)
+	return nil
+}
+
+// UnmarshalYAML is the custom unmarshal handler for this struct. It is
+// primarily useful for setting the defaults.
+func (obj *ConfigEtcdRes) UnmarshalYAML(unmarshal func(interface{}) error) error {
+	type rawRes ConfigEtcdRes // indirection to avoid infinite recursion
+
+	def := obj.Default()            // get the default
+	res, ok := def.(*ConfigEtcdRes) // put in the right format
+	if !ok {
+		return fmt.Errorf("could not convert to ConfigEtcdRes")
+	}
+	raw := rawRes(*res) // convert; the defaults go here
+
+	if err := unmarshal(&raw); err != nil {
+		return err
+	}
+
+	*obj = ConfigEtcdRes(raw) // restore from indirection with type conversion!
+	return nil
+}

engine/resources/consul_kv.go

@@ -0,0 +1,296 @@
+// Mgmt
+// Copyright (C) 2013-2024+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU General Public License for more details.
+//
+// You should have received a copy of the GNU General Public License
+// along with this program.  If not, see <https://www.gnu.org/licenses/>.
+//
+// Additional permission under GNU GPL version 3 section 7
+//
+// If you modify this program, or any covered work, by linking or combining it
+// with embedded mcl code and modules (and that the embedded mcl code and
+// modules which link with this program, contain a copy of their source code in
+// the authoritative form) containing parts covered by the terms of any other
+// license, the licensors of this program grant you additional permission to
+// convey the resulting work. Furthermore, the licensors of this program grant
+// the original author, James Shubin, additional permission to update this
+// additional permission if he deems it necessary to achieve the goals of this
+// additional permission.
+
+package resources
+
+import (
+	"context"
+	"fmt"
+	"net/url"
+	"sync"
+
+	"github.com/purpleidea/mgmt/engine"
+	"github.com/purpleidea/mgmt/engine/traits"
+	"github.com/purpleidea/mgmt/util"
+	"github.com/purpleidea/mgmt/util/errwrap"
+
+	"github.com/hashicorp/consul/api"
+)
+
+func init() {
+	engine.RegisterResource("consul:kv", func() engine.Res { return &ConsulKVRes{} })
+}
+
+// ConsulKVRes is a resource that writes a value into a Consul datastore. The
+// name of the resource can either be the key name, or the concatenation of the
+// server address and the key name: http://127.0.0.1:8500/my-key. If the param
+// keys are specified, then those are used. If the Name cannot be properly
+// parsed by url.Parse, then it will be considered as the Key's value. If the
+// Key is specified explicitly, then we won't use anything from the Name.
+type ConsulKVRes struct {
+	traits.Base
+	init *engine.Init
+
+	// Key is the name of the key. Defaults to the name of the resource.
+	Key string `lang:"key" yaml:"key"`
+
+	// Value is the value for the key.
+	Value string `lang:"value" yaml:"value"`
+
+	// Scheme is the URI scheme for the Consul server. Default: http.
+	Scheme string `lang:"scheme" yaml:"scheme"`
+
+	// Address is the address of the Consul server. Default: 127.0.0.1:8500.
+	Address string `lang:"address" yaml:"address"`
+
+	// Token is used to provide an ACL token to use for this resource.
+	Token string `lang:"token" yaml:"token"`
+
+	client *api.Client
+	config *api.Config // needed to close the idle connections
+	once   bool        // safety token
+	key    string      // cache the key name to avoid re-running the parser
+}
+
+// Default returns some sensible defaults for this resource.
+func (obj *ConsulKVRes) Default() engine.Res {
+	return &ConsulKVRes{}
+}
+
+// Validate if the params passed in are valid data.
+func (obj *ConsulKVRes) Validate() error {
+	s, _, k := obj.inputParser()
+	if k == "" {
+		return fmt.Errorf("the Key is empty")
+	}
+	if s != "" && s != "http" && s != "https" {
+		return fmt.Errorf("unknown Scheme")
+	}
+	return nil
+}
+
+// Init runs some startup code for this resource.
+func (obj *ConsulKVRes) Init(init *engine.Init) error {
+	obj.init = init // save for later
+
+	s, a, k := obj.inputParser()
+
+	obj.config = api.DefaultConfig()
+	if s != "" {
+		obj.config.Scheme = s
+	}
+	if a != "" {
+		obj.config.Address = obj.Address
+	}
+	obj.key = k // store the key
+	obj.init.Logf("using consul key: %s", obj.key)
+
+	if obj.Token != "" {
+		obj.config.Token = obj.Token
+	}
+
+	var err error
+	obj.client, err = api.NewClient(obj.config)
+	return errwrap.Wrapf(err, "could not create Consul client")
+}
+
+// Cleanup is run by the engine to clean up after the resource is done.
+func (obj *ConsulKVRes) Cleanup() error {
+	if obj.config != nil && obj.config.Transport != nil {
+		obj.config.Transport.CloseIdleConnections()
+	}
+	return nil
+}
+
+// Watch is the listener and main loop for this resource and it outputs events.
+func (obj *ConsulKVRes) Watch(ctx context.Context) error {
+	wg := &sync.WaitGroup{}
+	defer wg.Wait()
+
+	ch := make(chan error)
+	exit := make(chan struct{})
+
+	kv := obj.client.KV()
+
+	wg.Add(1)
+	go func() {
+		defer close(ch)
+		defer wg.Done()
+
+		opts := &api.QueryOptions{RequireConsistent: true}
+		innerCtx, cancel := util.ContextWithCloser(context.Background(), exit)
+		defer cancel()
+		opts = opts.WithContext(innerCtx)
+
+		for {
+			_, meta, err := kv.Get(obj.key, opts)
+			select {
+			case ch <- err: // send
+				if err != nil {
+					return
+				}
+
+				// WaitIndex = 0, which means that it is the
+				// first time we run the query, as we are about
+				// to change the WaitIndex to make a blocking
+				// query, we can consider the watch started.
+				opts.WaitIndex = meta.LastIndex
+				if opts.WaitIndex != 0 {
+					continue
+				}
+
+				if !obj.once {
+					obj.init.Running()
+					obj.once = true
+					continue
+				}
+
+				// Unexpected situation, bug in consul API...
+				select {
+				case ch <- fmt.Errorf("unexpected behaviour in Consul API"):
+				case <-ctx.Done(): // signal for shutdown request
+				}
+
+			case <-ctx.Done(): // signal for shutdown request
+			}
+			return
+		}
+	}()
+
+	defer close(exit)
+	for {
+		select {
+		case err, ok := <-ch:
+			if !ok { // channel shutdown
+				return nil
+			}
+			if err != nil {
+				return errwrap.Wrapf(err, "unknown %s watcher error", obj)
+			}
+			if obj.init.Debug {
+				obj.init.Logf("event!")
+			}
+			obj.init.Event()
+
+		case <-ctx.Done(): // signal for shutdown request
+			return nil
+		}
+	}
+}
+
+// CheckApply is run to check the state and, if apply is true, to apply the
+// necessary changes to reach the desired state. This is run before Watch and
+// again if Watch finds a change occurring to the state.
+func (obj *ConsulKVRes) CheckApply(ctx context.Context, apply bool) (bool, error) {
+	// XXX: use ctx for get and put
+	if obj.init.Debug {
+		obj.init.Logf("consul key: %s", obj.key)
+	}
+	kv := obj.client.KV()
+	pair, _, err := kv.Get(obj.key, nil)
+	if err != nil {
+		return false, err
+	}
+
+	if pair != nil && string(pair.Value) == obj.Value {
+		return true, nil
+	}
+
+	if !apply {
+		return false, nil
+	}
+
+	p := &api.KVPair{Key: obj.key, Value: []byte(obj.Value)}
+	_, err = kv.Put(p, nil)
+	return false, err
+}
+
+// Cmp compares two resources and return if they are equivalent.
+func (obj *ConsulKVRes) Cmp(r engine.Res) error {
+	res, ok := r.(*ConsulKVRes)
+	if !ok {
+		return fmt.Errorf("not a %s", obj.Kind())
+	}
+
+	if obj.Key != res.Key {
+		return fmt.Errorf("the Key param differs")
+	}
+	if obj.Value != res.Value {
+		return fmt.Errorf("the Value param differs")
+	}
+	if obj.Scheme != res.Scheme {
+		return fmt.Errorf("the Scheme param differs")
+	}
+	if obj.Address != res.Address {
+		return fmt.Errorf("the Address param differs")
+	}
+	if obj.Token != res.Token {
+		return fmt.Errorf("the Token param differs")
+	}
+
+	return nil
+}
+
+// inputParser parses the Name() of a resource and extracts the scheme, address,
+// and key name of a consul key. We don't have an error, because if we have one,
+// then it means the input must be a raw key. Output of this function is scheme,
+// address (includes hostname and port), and key. This also takes our parameters
+// in to account, and applies the correct overrides if they are specified there.
+func (obj *ConsulKVRes) inputParser() (string, string, string) {
+	// If the key is specified explicitly, then we're not going to parse the
+	// resource name for a pattern, and we use our given params as they are.
+	if obj.Key != "" {
+		return obj.Scheme, obj.Address, obj.Key
+	}
+
+	// Now we parse...
+	u, err := url.Parse(obj.Name())
+	if err != nil {
+		// If this didn't work, then we know it's explicitly a raw key.
+		return obj.Scheme, obj.Address, obj.Name()
+	}
+
+	// Otherwise, we use the parse result, and we overwrite any of the
+	// fields if we have an explicit param that was specified.
+	k := u.Path
+	s := u.Scheme
+	a := u.Host
+
+	//if obj.Key != "" { // this is now guaranteed to never happen
+	//	k = obj.Key
+	//}
+	if obj.Scheme != "" {
+		s = obj.Scheme
+	}
+	if obj.Address != "" {
+		a = obj.Address
+	}
+
+	return s, a, k
+}