Merge pull request #152 from runatlantis/atlantis-yaml-june

[WIP] Atlantis 0.4

.circleci/config.yml

@@ -3,7 +3,7 @@ jobs:
   test:
     working_directory: /go/src/github.com/runatlantis/atlantis
     docker:
-    - image: circleci/golang:1.10
+    - image: runatlantis/testing-env
     steps:
     - checkout
     - run: make test-coverage

.gitignore

@@ -8,3 +8,5 @@ website/src/public
 .DS_Store
 .cover
 .terraform/
+node_modules/
+**/.vuepress/dist

CHANGELOG.md

@@ -1,3 +1,23 @@
+# v0.4.0-alpha
+
+## Features
+* Autoplanning - Atlantis will automatically run `plan` on new pull requests and
+when new commits are pushed to the pull request.
+* New repository `atlantis.yaml` format that supports:
+    * Arbitrary step ordering
+    * Single config file for whole repository
+    * Controlling autoplanning
+* Moved docs to standalone website from the README.
+
+## Bugfixes
+
+## Backwards Incompatibilities / Notes:
+
+## Downloads
+
+## Docker
+
+
 # v0.3.10
 
 ## Features

Gopkg.toml

@@ -88,3 +88,15 @@
 [[constraint]]
   branch = "master"
   name = "github.com/lkysow/go-gitlab"
+
+[[constraint]]
+  name = "github.com/go-test/deep"
+  version = "1.0.1"
+
+[[constraint]]
+  branch = "master"
+  name = "github.com/flynn-archive/go-shlex"
+
+[[constraint]]
+  branch = "master"
+  name = "github.com/docker/docker"

Makefile

@@ -28,13 +28,17 @@ build-service: ## Build the main Go service
 go-generate: ## Run go generate in all packages
 	go generate $(PKG)
 
-regen-mocks: ## Delete all mocks and then run go generate to regen them
-	find . -type f | grep mocks/mock | grep -v vendor | xargs rm
-	@# not using $(PKG) here because that it includes directories that have now
-	@# been deleted, causing go generate to fail.
-	go generate $$(go list ./... | grep -v e2e | grep -v vendor | grep -v static)
+#regen-mocks: ## Delete all mocks and matchers and then run go generate to regen them. This doesn't work anymore.
+#find . -type f | grep mocks/mock_ | grep -v vendor | xargs rm
+#find . -type f | grep mocks/matchers | grep -v vendor | xargs rm
+#@# not using $(PKG) here because that it includes directories that have now
+#@# been deleted, causing go generate to fail.
+#echo "this doesn't work anymore: go generate \$\$(go list ./... | grep -v e2e | grep -v vendor | grep -v static)"
 
 test: ## Run tests
+	@go test -short $(PKG)
+
+test-all: ## Run tests including integration
 	@go test $(PKG)
 
 test-coverage:
@@ -78,9 +82,9 @@ end-to-end-tests: ## Run e2e tests
 	./scripts/e2e.sh
 
 generate-website-html: ## Generate HTML for website
-	cd website/src && hugo -d ../html
+	yarn website:build
 
 upload-website-html: ## Upload generated website to s3
 	aws s3 rm s3://www.runatlantis.io/ --recursive
-	aws s3 sync website/html/ s3://www.runatlantis.io/
-	rm -rf website/html/
+	aws s3 sync runatlantis.io/.vuepress/dist/ s3://www.runatlantis.io/
+	rm -rf runatlantis.io/.vuepress/dist

README.md

@@ -124,42 +124,42 @@ Atlantis supports several Terraform project structures:
 ```
 .
 ├── project1
-│   ├── main.tf
+│   ├── main.tf
 |   └── ...
 └── project2
-    ├── main.tf
+    ├── main.tf
     └── ...
 ```
 -  one folder per set of configuration
 ```
 .
 ├── staging
-│   ├── main.tf
+│   ├── main.tf
 |   └── ...
 └── production
-    ├── main.tf
+    ├── main.tf
     └── ...
 ```
 -  using `env/{env}.tfvars` to define workspace specific variables. This works in both multi-project repos and single-project repos.
 ```
 .
 ├── env
-│   ├── production.tfvars
-│   └── staging.tfvars
+│   ├── production.tfvars
+│   └── staging.tfvars
 └── main.tf
 ```
 or
 ```
 .
 ├── project1
-│   ├── env
-│   │   ├── production.tfvars
-│   │   └── staging.tfvars
-│   └── main.tf
+│   ├── env
+│   │   ├── production.tfvars
+│   │   └── staging.tfvars
+│   └── main.tf
 └── project2
     ├── env
-    │   ├── production.tfvars
-    │   └── staging.tfvars
+    │   ├── production.tfvars
+    │   └── staging.tfvars
     └── main.tf
 ```
 With the above project structure you can de-duplicate your Terraform code between workspaces/environments without requiring extensive use of modules. At Hootsuite we found this project format to be very successful and use it in all of our 100+ Terraform repositories.

cmd/server.go

@@ -32,8 +32,9 @@ import (
 // 3. Add your flag's description etc. to the stringFlags, intFlags, or boolFlags slices.
 const (
 	// Flag names.
-	AtlantisURLFlag     = "atlantis-url"
 	AllowForkPRsFlag    = "allow-fork-prs"
+	AllowRepoConfigFlag = "allow-repo-config"
+	AtlantisURLFlag     = "atlantis-url"
 	ConfigFlag          = "config"
 	DataDirFlag         = "data-dir"
 	GHHostnameFlag      = "gh-hostname"
@@ -142,6 +143,13 @@ var boolFlags = []boolFlag{
 		description:  "Allow Atlantis to run on pull requests from forks. A security issue for public repos.",
 		defaultValue: false,
 	},
+	{
+		name: AllowRepoConfigFlag,
+		description: "Allow repositories to use atlantis.yaml files to customize the commands Atlantis runs." +
+			" Should only be enabled in a trusted environment since it enables a pull request to run arbitrary commands" +
+			" on the Atlantis server.",
+		defaultValue: false,
+	},
 	{
 		name:         RequireApprovalFlag,
 		description:  "Require pull requests to be \"Approved\" before allowing the apply command to be run.",
@@ -298,8 +306,9 @@ func (s *ServerCmd) run() error {
 
 	// Config looks good. Start the server.
 	server, err := s.ServerCreator.NewServer(userConfig, server.Config{
-		AllowForkPRsFlag: AllowForkPRsFlag,
-		AtlantisVersion:  s.AtlantisVersion,
+		AllowForkPRsFlag:    AllowForkPRsFlag,
+		AllowRepoConfigFlag: AllowRepoConfigFlag,
+		AtlantisVersion:     s.AtlantisVersion,
 	})
 	if err != nil {
 		return errors.Wrap(err, "initializing server")