angular-cn/.circleci/config.yml

902 lines
37 KiB
YAML
Raw Normal View History

# Configuration file for https://circleci.com/gh/angular/angular
# Note: YAML anchors allow an object to be re-used, reducing duplication.
# The ampersand declares an alias for an object, then later the `<<: *name`
# syntax dereferences it.
# See http://blog.daemonl.com/2016/02/yaml.html
# To validate changes, use an online parser, eg.
# http://yaml-online-parser.appspot.com/
# Note that the browser docker image comes with Chrome and Firefox preinstalled. This is just
# needed for jobs that run tests without Bazel. Bazel runs tests with browsers that will be
# fetched by the Webtesting rules. Therefore for jobs that run tests with Bazel, we don't need a
# docker image with browsers pre-installed.
ci: pin docker images by ID for hermeticity (#32602) Previously, the docker images used on CI where specified by a tag (`10.16` and `10.16-browsers`). Since tags are not immutable, this only pins specific characteristics of the environment (e.g. the OS type and the Node.js version), but not others. Especially when using a tag that does not specify the patch version (e.g. `10.16` instead of `10.16.0`), it is inevitable that the image will change at some point, potentially leading to unrelated failures due to changes in the environment. One source of such failures can be the Chrome version used in tests. Since we install a specific ChromeDriver version (that is only compatible with specific Chrome version ranges), unexpectedly updating to a newer Chrome version may break the tests if the new version falls outside the range of supported version for our pinned ChromeDriver. Using a tag that specifies the patch version (e.g. `10.16.0`) or even the OS version (e.g. `10.16.0-buster`) is safer (i.e. has a lower probability of introducing the kind of breakages described above), but is still not fully hermetic. This commit prevents such breakages by pinning the docker images by ID. Image IDs are based on the image's digest (SHA256) and are thus immutable, ensuring that all CI jobs will be running on the exact same image. See [here][1] for more info on pre-built CircleCI docker images and more specifically [pinning images by ID][2]. [1]: https://circleci.com/docs/2.0/circleci-images [2]: https://circleci.com/docs/2.0/circleci-images#using-a-docker-image-id-to-pin-an-image-to-a-fixed-version PR Close #32602
2019-09-11 05:09:47 -04:00
# **NOTE 1**: Pin to exact images using an ID (SHA). See https://circleci.com/docs/2.0/circleci-images/#using-a-docker-image-id-to-pin-an-image-to-a-fixed-version.
# (Using the tag in not necessary when pinning by ID, but include it anyway for documentation purposes.)
# **NOTE 2**: If you change the version of the docker images, also change the `cache_key` suffix.
# **NOTE 3**: If you change the version of the `*-browsers` docker image, make sure the
# `CI_CHROMEDRIVER_VERSION_ARG` env var (in `.circleci/env.sh`) points to a ChromeDriver
# version that is compatible with the Chrome version in the image.
ci: pin docker images by ID for hermeticity (#32602) Previously, the docker images used on CI where specified by a tag (`10.16` and `10.16-browsers`). Since tags are not immutable, this only pins specific characteristics of the environment (e.g. the OS type and the Node.js version), but not others. Especially when using a tag that does not specify the patch version (e.g. `10.16` instead of `10.16.0`), it is inevitable that the image will change at some point, potentially leading to unrelated failures due to changes in the environment. One source of such failures can be the Chrome version used in tests. Since we install a specific ChromeDriver version (that is only compatible with specific Chrome version ranges), unexpectedly updating to a newer Chrome version may break the tests if the new version falls outside the range of supported version for our pinned ChromeDriver. Using a tag that specifies the patch version (e.g. `10.16.0`) or even the OS version (e.g. `10.16.0-buster`) is safer (i.e. has a lower probability of introducing the kind of breakages described above), but is still not fully hermetic. This commit prevents such breakages by pinning the docker images by ID. Image IDs are based on the image's digest (SHA256) and are thus immutable, ensuring that all CI jobs will be running on the exact same image. See [here][1] for more info on pre-built CircleCI docker images and more specifically [pinning images by ID][2]. [1]: https://circleci.com/docs/2.0/circleci-images [2]: https://circleci.com/docs/2.0/circleci-images#using-a-docker-image-id-to-pin-an-image-to-a-fixed-version PR Close #32602
2019-09-11 05:09:47 -04:00
var_1: &default_docker_image circleci/node:10.16@sha256:75c05084fff4afa3683a03c5a04a4a3ad95c536ff2439d8fe14e7e1f5c58b09a
var_2: &browsers_docker_image circleci/node:10.16-browsers@sha256:d2a96fe1cbef51257ee626b5f645e64dade3e886f00ba9cb7e8ea65b4efe8db1
# We don't want to include the current branch name in the cache key because that would prevent
# PRs from being able to restore the cache since the branch names are always different for PRs.
# The cache key should only consist of dynamic values that change whenever something in the
# cache changes. For example:
# 1) yarn lock file changes --> cached "node_modules" are different.
# 2) bazel repository definitions change --> cached bazel repositories are different.
# **NOTE 1 **: If you change the cache key prefix, also sync the restore_cache fallback to match.
# **NOTE 2 **: Keep the static part of the cache key as prefix to enable correct fallbacks.
# See https://circleci.com/docs/2.0/caching/#restoring-cache for how prefixes work in CircleCI.
var_3: &cache_key v3-angular-node-10.16-{{ checksum "yarn.lock" }}-{{ checksum "WORKSPACE" }}-{{ checksum "packages/bazel/package.bzl" }}-{{ checksum "aio/yarn.lock" }}
var_3_win: &cache_key_win v4-angular-win-node-12.0-{{ checksum "yarn.lock" }}-{{ checksum "WORKSPACE" }}-{{ checksum "packages/bazel/package.bzl" }}-{{ checksum "aio/yarn.lock" }}
# Initializes the CI environment by setting up common environment variables.
var_4: &init_environment
run:
name: Initializing environment (setting up variables, overwriting Yarn)
# Overwrite the yarn installed in the docker container with our own version.
command: |
./.circleci/env.sh
ourYarn=$(realpath ./third_party/github.com/yarnpkg/yarn/releases/download/v1.17.3/bin/yarn.js)
sudo chmod a+x $ourYarn
sudo ln -fs $ourYarn /usr/local/bin/yarn
echo "Yarn version: $(yarn --version)"
# Add GitHub to known hosts.
mkdir -p ~/.ssh
echo 'github.com ssh-rsa AAAAB3NzaC1yc2EAAAABIwAAAQEAq2A7hRGmdnm9tUDbO9IDSwBK6TbQa+PXYPCPy6rbTrTtw7PHkccKrpp0yVhp5HdEIcKr6pLlVDBfOLX9QUsyCOV0wzfjIJNlGEYsdlLJizHhbn2mUjvSAHQqZETYP81eFzLQNnPHt4EVVUh7VfDESU84KezmD5QlWpXLmvU31/yMf+Se8xhHTvKSCZIFImWwoG6mbUoWf9nzpIoaSjB+weqqUUmpaaasXVal72J+UX2B+2RPW3RcT0eOzQgqlJL3RKrTJvdsjE3JEAvGq3lGHSZXy28G3skua2SmVi/w4yCE6gbODqnTWlg7+wC604ydGXA8VJiS5ap43JXiUFFAaQ==' >> ~/.ssh/known_hosts
# use git+ssh instead of https
git config --global url."ssh://git@github.com".insteadOf "https://github.com" || true
git config --global gc.auto 0 || true
var_4_win: &init_environment_win
run:
# Install Bazel pre-requisites that aren't in the preconfigured CircleCI Windows VM.
name: Setup windows node environment
command: ./.circleci/windows-env.ps1
var_5: &setup_bazel_remote_execution
run:
name: "Setup bazel RBE remote execution"
command: |
# We need ensure that the same default digest is used for encoding and decoding
# with openssl. Openssl versions might have different default digests which can
# cause decryption failures based on the openssl version. https://stackoverflow.com/a/39641378/4317734
openssl aes-256-cbc -d -in .circleci/gcp_token -md md5 -k "$CI_REPO_NAME" -out /home/circleci/.gcp_credentials
echo "export GOOGLE_APPLICATION_CREDENTIALS=/home/circleci/.gcp_credentials" >> $BASH_ENV
./.circleci/setup-rbe.sh .bazelrc.user
# Settings common to each job
var_6: &job_defaults
working_directory: ~/ng
docker:
- image: *default_docker_image
var_6_win: &job_defaults_win
working_directory: ~/ng
resource_class: windows.medium
# CircleCI windows VMs do have the GitBash shell available:
# https://github.com/CircleCI-Public/windows-preview-docs#shells
# But in this specific case we really should not use it because Bazel must not be ran from
# GitBash. These issues discuss why:
# https://github.com/bazelbuild/bazel/issues/5751
# https://github.com/bazelbuild/bazel/issues/5724#issuecomment-410194038
# https://github.com/bazelbuild/bazel/issues/6339#issuecomment-441600879
shell: powershell.exe -ExecutionPolicy Bypass
machine:
# Preview image that includes the following:
# - Visual Studio 2019 build tools
# - Node 12
# - yarn 1.17
# - Python 3 3.7.4
image: windows-server-2019-vs2019:201908-02
# After checkout, rebase on top of target branch.
var_7: &post_checkout
run:
name: Rebase PR on target branch
command: >
if [[ -n "${CIRCLE_PR_NUMBER}" ]]; then
# User is required for rebase.
git config user.name "angular-ci"
git config user.email "angular-ci"
# Rebase PR on top of target branch.
node tools/rebase-pr.js angular/angular ${CIRCLE_PR_NUMBER}
else
echo "This build is not over a PR, nothing to do."
fi
var_7_win: &post_checkout_win
run:
name: Rebase PR on target branch
command: >
if (Test-Path env:CIRCLE_PR_NUMBER) {
git config user.name "angular-ci"
git config user.email "angular-ci"
node tools\rebase-pr.js angular/angular-cli $env:CIRCLE_PR_NUMBER
} else {
echo "This build is not over a PR, nothing to do."
}
var_8: &yarn_install
run:
name: Running Yarn install
command: |
# Yarn's requests sometimes take more than 10mins to complete.
# Print something to stdout, to prevent CircleCI from failing due to not output.
while true; do sleep 60; echo "[`date`] Keeping alive..."; done &
KEEP_ALIVE_PID=$!
yarn install --frozen-lockfile --non-interactive
kill $KEEP_ALIVE_PID
var_9: &setup_circleci_bazel_config
run:
name: Setting up CircleCI bazel configuration
command: sudo cp .circleci/bazel.linux.rc $HOME/.bazelrc
var_9_win: &setup_circleci_bazel_config_win
run:
name: Setting up CircleCI bazel configuration
command: copy .circleci\bazel.windows.rc $env:USERPROFILE\.bazelrc
var_10: &restore_cache
restore_cache:
keys:
- *cache_key
# This fallback should be the cache_key without variables.
- v3-angular-node-10.16-
var_10_win: &restore_cache_win
restore_cache:
keys:
- *cache_key_win
# This fallback should be the cache_key without variables.
- v4-angular-win-node-12.0-
# Branch filter that can be specified for jobs that should only run on publish branches
# (e.g. master or the patch branch)
var_11: &publish_branches_filter
branches:
only:
- master
# e.g. 7.0.x, 7.1.x, etc.
- /\d+\.\d+\.x/
# Workspace initially persisted by the `install` job, and then enhanced by `test_aio` and
# `build-npm-packages`.
# https://circleci.com/docs/2.0/workflows/#using-workspaces-to-share-data-among-jobs
# https://circleci.com/blog/deep-diving-into-circleci-workspaces/
var_12: &attach_workspace
attach_workspace:
at: ~/
var_13: &notify_caretaker_on_fail
run:
when: on_fail
name: Notify caretaker about failure
# `$SLACK_CARETAKER_WEBHOOK_URL` is a secret env var defined in CircleCI project settings.
# The URL comes from https://angular-team.slack.com/apps/A0F7VRE7N-circleci.
command: |
notificationJson="{\"text\":\":x: \`$CIRCLE_JOB\` job for $CIRCLE_BRANCH branch failed on build $CIRCLE_BUILD_NUM: $CIRCLE_BUILD_URL :scream:\"}"
curl --request POST --header "Content-Type: application/json" --data "$notificationJson" $SLACK_CARETAKER_WEBHOOK_URL
var_14: &notify_dev_infra_on_fail
run:
when: on_fail
name: Notify dev-infra about failure
# `$SLACK_DEV_INFRA_CI_FAILURES_WEBHOOK_URL` is a secret env var defined in CircleCI project settings.
# The URL comes from https://angular-team.slack.com/apps/A0F7VRE7N-circleci.
command: |
notificationJson="{\"text\":\":x: \`$CIRCLE_JOB\` job for $CIRCLE_BRANCH branch failed on build $CIRCLE_BUILD_NUM: $CIRCLE_BUILD_URL :scream:\"}"
curl --request POST --header "Content-Type: application/json" --data "$notificationJson" $SLACK_DEV_INFRA_CI_FAILURES_WEBHOOK_URL
# Cache key for the Material unit tests job. **Note** when updating the SHA in the cache keys,
# also update the SHA for the "MATERIAL_REPO_COMMIT" environment variable.
var_15: &material_unit_tests_cache_key v4-angular-material-2817c9e2faa4140342336987a692d5dd30bf24c2
var_16: &material_unit_tests_cache_key_short v4-angular-material
version: 2
jobs:
setup:
<<: *job_defaults
steps:
- checkout
- *post_checkout
# This cache is saved in the build-npm-packages so that Bazel cache is also included.
- *restore_cache
- *init_environment
- *yarn_install
- run: yarn --cwd aio install --frozen-lockfile --non-interactive
# Make the bazel directories and add a file to them if they don't exist already so that
# persist_to_workspace does not fail.
- run: |
if [ ! -d ~/bazel_repository_cache ]; then
mkdir ~/bazel_repository_cache
touch ~/bazel_repository_cache/MARKER
fi
# Persist any changes at this point to be reused by further jobs.
# **NOTE 1 **: Folders persisted here should be kept in sync with `var_13: &attach_workspace`.
# **NOTE 2 **: To add new content to the workspace, always persist on the same root.
- persist_to_workspace:
root: ~/
paths:
- ./ng
- ./bazel_repository_cache
lint:
<<: *job_defaults
steps:
- *attach_workspace
- *init_environment
- run: 'yarn bazel:format -mode=check ||
(echo "BUILD files not formatted. Please run ''yarn bazel:format''" ; exit 1)'
# Run the skylark linter to check our Bazel rules
- run: 'yarn bazel:lint ||
(echo -e "\n.bzl files have lint errors. Please run ''yarn bazel:lint-fix''"; exit 1)'
- run: yarn gulp lint
- run: node tools/verify-codeownership
test:
<<: *job_defaults
resource_class: xlarge
steps:
- *attach_workspace
- *init_environment
- *setup_circleci_bazel_config
# Setup remote execution and run RBE-compatible tests.
- *setup_bazel_remote_execution
- run:
command: yarn bazel test //... --build_tag_filters=-ivy-only --test_tag_filters=-ivy-only
no_output_timeout: 20m
# Temporary job to test what will happen when we flip the Ivy flag to true
test_ivy_aot:
<<: *job_defaults
resource_class: xlarge
steps:
- *attach_workspace
- *init_environment
- *setup_circleci_bazel_config
- *setup_bazel_remote_execution
# We need to explicitly specify the --symlink_prefix option because otherwise we would
# not be able to easily find the output bin directory when uploading artifacts for size
# measurements.
- run:
command: yarn test-ivy-aot //... --symlink_prefix=dist/
no_output_timeout: 20m
# Publish bundle artifacts which will be used to calculate the size change. **Note**: Make
# sure that the size plugin from the Angular robot fetches the artifacts from this CircleCI
# job (see .github/angular-robot.yml). Additionally any artifacts need to be stored with the
# following path format: "{projectName}/{context}/{fileName}". This format is necessary
# because otherwise the bot is not able to pick up the artifacts from CircleCI. See:
# https://github.com/angular/github-robot/blob/master/functions/src/plugins/size.ts#L392-L394
- store_artifacts:
path: dist/bin/packages/core/test/bundling/hello_world/bundle.min.js
destination: core/hello_world/bundle
- store_artifacts:
path: dist/bin/packages/core/test/bundling/todo/bundle.min.js
destination: core/todo/bundle
- store_artifacts:
path: dist/bin/packages/core/test/bundling/hello_world/bundle.min.js.br
destination: core/hello_world/bundle.br
- store_artifacts:
path: dist/bin/packages/core/test/bundling/todo/bundle.min.js.br
destination: core/todo/bundle.br
# This job is currently a PoC for running tests on SauceLabs via bazel. It runs a subset of the
# tests in `legacy-unit-tests-saucelabs` (see
# [BUILD.bazel](https://github.com/angular/angular/blob/ef44f51d5/BUILD.bazel#L66-L92)).
#
# NOTE: This is currently limited to master builds only. See the `default_workflow` configuration.
test_saucelabs_bazel:
<<: *job_defaults
# In order to avoid the bottleneck of having a slow host machine, we acquire a better
# container for this job. This is necessary because we launch a lot of browsers concurrently
# and therefore the tunnel and Karma need to process a lot of file requests and tests.
resource_class: xlarge
steps:
- *attach_workspace
- *init_environment
- *setup_circleci_bazel_config
- run:
name: Run Bazel tests in saucelabs
# All web tests are contained within a single //:test_web_all target for Saucelabs
# as running each set of tests as a separate target will attempt to acquire too
# many browsers on Saucelabs (7 per target currently) and some tests will always
# fail to acquire browsers. For example:
# 14 02 2019 19:52:33.170:WARN [launcher]: chrome beta on SauceLabs have not captured in 180000 ms, killing.
# //packages/forms/test:web_test_sauce TIMEOUT in 315.0s
command: |
./scripts/saucelabs/run-bazel-via-tunnel.sh \
--tunnel-id angular-${CIRCLE_BUILD_NUM}-${CIRCLE_NODE_INDEX} \
--username $SAUCE_USERNAME \
--key $(echo $SAUCE_ACCESS_KEY | rev) \
yarn bazel test //:test_web_all
no_output_timeout: 20m
- *notify_dev_infra_on_fail
test_aio:
<<: *job_defaults
docker:
# Needed because the AIO tests and the PWA score test depend on Chrome being available.
- image: *browsers_docker_image
steps:
- *attach_workspace
- *init_environment
# Build aio
- run: yarn --cwd aio build --progress=false
# Lint the code
- run: yarn --cwd aio lint
# Run unit tests
- run: yarn --cwd aio test --progress=false --watch=false
# Run e2e tests
- run: yarn --cwd aio e2e --configuration=ci
# Run PWA-score tests
- run: yarn --cwd aio test-pwa-score-localhost $CI_AIO_MIN_PWA_SCORE
# Run accessibility tests
- run: yarn --cwd aio test-a11y-score-localhost
# Check the bundle sizes.
- run: yarn --cwd aio payload-size
# Run unit tests for Firebase redirects
- run: yarn --cwd aio redirects-test
deploy_aio:
<<: *job_defaults
docker:
# Needed because before deploying the deploy-production script runs the PWA score tests.
- image: *browsers_docker_image
steps:
- *attach_workspace
- *init_environment
# Deploy angular.io to production (if necessary)
- run: setPublicVar_CI_STABLE_BRANCH
- run: yarn --cwd aio deploy-production
test_aio_local:
<<: *job_defaults
docker:
# Needed because the AIO tests and the PWA score test depend on Chrome being available.
- image: *browsers_docker_image
steps:
- *attach_workspace
- *init_environment
# Build aio (with local Angular packages)
- run: yarn --cwd aio build-local-ci
# Run unit tests
- run: yarn --cwd aio test --progress=false --watch=false
# Run e2e tests
- run: yarn --cwd aio e2e --configuration=ci
# Run PWA-score tests
- run: yarn --cwd aio test-pwa-score-localhost $CI_AIO_MIN_PWA_SCORE
# Check the bundle sizes.
- run: yarn --cwd aio payload-size aio-local
test_aio_local_ivy:
<<: *job_defaults
docker:
# Needed because the AIO tests and the PWA score test depend on Chrome being available.
- image: *browsers_docker_image
steps:
- *attach_workspace
- *init_environment
# Build aio with Ivy (using local Angular packages)
- run: yarn --cwd aio build-with-ivy-ci
# Run unit tests
- run: yarn --cwd aio test --progress=false --watch=false
# Run e2e tests
- run: yarn --cwd aio e2e --configuration=ci
# Run PWA-score tests
- run: yarn --cwd aio test-pwa-score-localhost $CI_AIO_MIN_PWA_SCORE
# Check the bundle sizes.
- run: yarn --cwd aio payload-size aio-local-ivy
test_aio_tools:
<<: *job_defaults
steps:
- *attach_workspace
- *init_environment
# Install
- run: yarn --cwd aio install --frozen-lockfile --non-interactive
- run: yarn --cwd aio extract-cli-command-docs
# Run tools tests
- run: yarn --cwd aio tools-test
- run: ./aio/aio-builds-setup/scripts/test.sh
test_docs_examples:
<<: *job_defaults
docker:
# Needed because the example e2e tests depend on Chrome.
- image: *browsers_docker_image
parallelism: 4
resource_class: xlarge
steps:
- *attach_workspace
- *init_environment
# Install aio
- run: yarn --cwd aio install --frozen-lockfile --non-interactive
# Run examples tests. The "CIRCLE_NODE_INDEX" will be set if "parallelism" is enabled.
# Since the parallelism is set to "3", there will be three parallel CircleCI containers
# with either "0", "1" or "2" as node index. This can be passed to the "--shard" argument.
- run: yarn --cwd aio example-e2e --setup --local --cliSpecsConcurrency=5 --shard=${CIRCLE_NODE_INDEX}/${CIRCLE_NODE_TOTAL} --retry 2
test_docs_examples_ivy:
<<: *job_defaults
docker:
# Needed because the example e2e tests depend on Chrome.
- image: *browsers_docker_image
resource_class: xlarge
# We increase the parallelism here to five while the "test_docs_examples" job runs with
# a parallelism of four. This is necessary because this job also need to run NGCC which
# takes up more time and we don't want these jobs to impact the overall CI turnaround.
parallelism: 5
steps:
- *attach_workspace
- *init_environment
# Install aio
- run: yarn --cwd aio install --frozen-lockfile --non-interactive
# Rename the Ivy packages dist folder to "dist/packages-dist" as the AIO
# package installer picks up the locally built packages from that location.
# *Note*: We could also adjust the packages installer, but given we won't have
# two different folders of Angular distributions in the future, we should keep
# the packages installer unchanged.
- run: mv dist/packages-dist-ivy-aot dist/packages-dist
# Run examples tests with ivy. The "CIRCLE_NODE_INDEX" will be set if "parallelism" is enabled.
# Since the parallelism is set to "3", there will be three parallel CircleCI containers
# with either "0", "1" or "2" as node index. This can be passed to the "--shard" argument.
- run: yarn --cwd aio example-e2e --setup --local --ivy --cliSpecsConcurrency=5 --shard=${CIRCLE_NODE_INDEX}/${CIRCLE_NODE_TOTAL} --retry 2
# This job should only be run on PR builds, where `CI_PULL_REQUEST` is not `false`.
aio_preview:
<<: *job_defaults
environment:
AIO_SNAPSHOT_ARTIFACT_PATH: &aio_preview_artifact_path 'aio/tmp/snapshot.tgz'
steps:
- *attach_workspace
- *init_environment
- run: ./aio/scripts/build-artifacts.sh $AIO_SNAPSHOT_ARTIFACT_PATH $CI_PULL_REQUEST $CI_COMMIT
- store_artifacts:
path: *aio_preview_artifact_path
# The `destination` needs to be kept in synch with the value of
# `AIO_ARTIFACT_PATH` in `aio/aio-builds-setup/Dockerfile`
destination: aio/dist/aio-snapshot.tgz
- run: node ./aio/scripts/create-preview $CIRCLE_BUILD_NUM
# This job should only be run on PR builds, where `CI_PULL_REQUEST` is not `false`.
test_aio_preview:
<<: *job_defaults
docker:
# Needed because the test-preview script runs e2e tests and the PWA score test with Chrome.
- image: *browsers_docker_image
steps:
- *attach_workspace
- *init_environment
- run: yarn --cwd aio install --frozen-lockfile --non-interactive
- run:
name: Wait for preview and run tests
command: node aio/scripts/test-preview.js $CI_PULL_REQUEST $CI_COMMIT $CI_AIO_MIN_PWA_SCORE
# The `build-npm-packages` tasks exist for backwards-compatibility with old scripts and
# tests that rely on the pre-Bazel `dist/packages-dist` output structure (build.sh).
# Having multiple jobs that independently build in this manner duplicates some work; we build
# the bazel packages more than once. Even though we have a remote cache, these jobs will
# typically run in parallel so up-to-date outputs will not be available at the time the build
# starts.
# Build the view engine npm packages. No new jobs should depend on this.
build-npm-packages:
<<: *job_defaults
resource_class: xlarge
steps:
- *attach_workspace
- *init_environment
- *setup_circleci_bazel_config
- *setup_bazel_remote_execution
- run: scripts/build-packages-dist.sh
# Save the npm packages from //packages/... for other workflow jobs to read
- persist_to_workspace:
root: ~/
paths:
- ng/dist/packages-dist
# Save dependencies and bazel repository cache to use on subsequent runs.
- save_cache:
key: *cache_key
paths:
- "node_modules"
- "aio/node_modules"
- "~/bazel_repository_cache"
# Build the ivy npm packages.
build-ivy-npm-packages:
<<: *job_defaults
resource_class: xlarge
steps:
- *attach_workspace
- *init_environment
- *setup_circleci_bazel_config
- *setup_bazel_remote_execution
- run: scripts/build-ivy-npm-packages.sh
# Save the npm packages from //packages/... for other workflow jobs to read
- persist_to_workspace:
root: ~/
paths:
- ng/dist/packages-dist-ivy-aot
# We run the integration tests outside of Bazel for now.
# They are a separate workflow job so that they can be easily re-run.
# When the tests are ported to bazel test targets, they should move to the "test"
# job above, as part of the bazel test command. That has flaky_test_attempts so the
# need to re-run manually should be alleviated.
# See comments inside the integration/run_tests.sh script.
integration_test:
<<: *job_defaults
parallelism: 4
docker:
# Needed because the integration tests expect Chrome to be installed (e.g cli-hello-world)
- image: *browsers_docker_image
# Note: we run Bazel in one of the integration tests, and it can consume >2G
# of memory. Together with the system under test, this can exhaust the RAM
# on a 4G worker so we use a larger machine here too.
resource_class: xlarge
steps:
- *attach_workspace
- *init_environment
# Runs the integration tests in parallel across multiple CircleCI container instances. The
# amount of container nodes for this job is controlled by the "parallelism" option.
- run: ./integration/run_tests.sh ${CIRCLE_NODE_INDEX} ${CIRCLE_NODE_TOTAL}
# This job updates the content of repos like github.com/angular/core-builds
# for every green build on angular/angular.
publish_snapshot:
<<: *job_defaults
steps:
# See below - ideally this job should not trigger for non-upstream builds.
# But since it does, we have to check this condition.
- run:
name: Skip this job for Pull Requests and Fork builds
# Note: Using `CIRCLE_*` env variables (instead of those defined in `env.sh` so that this
# step can be run before `init_environment`.
command: >
if [[ -n "${CIRCLE_PR_NUMBER}" ]] ||
[[ "$CIRCLE_PROJECT_USERNAME" != "angular" ]] ||
[[ "$CIRCLE_PROJECT_REPONAME" != "angular" ]]; then
circleci step halt
fi
- *attach_workspace
- *init_environment
# CircleCI has a config setting to force SSH for all github connections
# This is not compatible with our mechanism of using a Personal Access Token
# Clear the global setting
- run: git config --global --unset "url.ssh://git@github.com.insteadof"
- run:
name: Decrypt github credentials
# We need ensure that the same default digest is used for encoding and decoding with
# openssl. Openssl versions might have different default digests which can cause
# decryption failures based on the installed openssl version. https://stackoverflow.com/a/39641378/4317734
command: 'openssl aes-256-cbc -d -in .circleci/github_token -md md5 -k "${KEY}" -out ~/.git_credentials'
- run: ./scripts/ci/publish-build-artifacts.sh
aio_monitoring_stable:
<<: *job_defaults
docker:
# This job needs Chrome to be globally installed because the tests run with Protractor
# which does not load the browser through the Bazel webtesting rules.
- image: *browsers_docker_image
steps:
- *attach_workspace
- *init_environment
- run: setPublicVar_CI_STABLE_BRANCH
- run:
name: Check out `aio/` from the stable branch
command: |
git fetch origin $CI_STABLE_BRANCH
git checkout --force origin/$CI_STABLE_BRANCH -- aio/
- run:
name: Run tests against https://angular.io/
command: ./aio/scripts/test-production.sh https://angular.io/ $CI_AIO_MIN_PWA_SCORE
- *notify_caretaker_on_fail
- *notify_dev_infra_on_fail
aio_monitoring_next:
<<: *job_defaults
docker:
# This job needs Chrome to be globally installed because the tests run with Protractor
# which does not load the browser through the Bazel webtesting rules.
- image: *browsers_docker_image
steps:
- *attach_workspace
- *init_environment
- run:
name: Run tests against https://next.angular.io/
command: ./aio/scripts/test-production.sh https://next.angular.io/ $CI_AIO_MIN_PWA_SCORE
- *notify_caretaker_on_fail
- *notify_dev_infra_on_fail
legacy-unit-tests-saucelabs:
<<: *job_defaults
# In order to avoid the bottleneck of having a slow host machine, we acquire a better
# container for this job. This is necessary because we launch a lot of browsers concurrently
# and therefore the tunnel and Karma need to process a lot of file requests and tests.
resource_class: xlarge
steps:
- *attach_workspace
- *init_environment
- run:
name: Preparing environment for running tests on Saucelabs.
command: |
setPublicVar KARMA_JS_BROWSERS $(node -e 'console.log(require("./browser-providers.conf").sauceAliases.CI_REQUIRED.join(","))')
setSecretVar SAUCE_ACCESS_KEY $(echo $SAUCE_ACCESS_KEY | rev)
- run:
name: Starting Saucelabs tunnel
command: ./scripts/saucelabs/start-tunnel.sh
background: true
- run: yarn tsc -p packages
- run: yarn tsc -p modules
# Waits for the Saucelabs tunnel to be ready. This ensures that we don't run tests
# too early without Saucelabs not being ready.
- run: ./scripts/saucelabs/wait-for-tunnel.sh
- run: yarn karma start ./karma-js.conf.js --single-run --browsers=${KARMA_JS_BROWSERS}
- run: ./scripts/saucelabs/stop-tunnel.sh
legacy-misc-tests:
<<: *job_defaults
steps:
- *attach_workspace
- *init_environment
- run: yarn gulp check-cycle
# TODO: disabled because the Bazel packages-dist does not seem to have map files for
# the ESM5/ES2015 output. See: https://github.com/angular/angular/issues/27966
# - run: yarn gulp source-map-test
# Job to run unit tests from angular/material2. Needs a browser since all
# component unit tests assume they're running in the browser environment.
material-unit-tests:
<<: *job_defaults
resource_class: xlarge
docker:
- image: *browsers_docker_image
steps:
- *attach_workspace
- *init_environment
# Although RBE is configured below for the Material repo, also setup RBE in the Angular repo
# to provision Angular's GCP token into the environment variables.
- *setup_bazel_remote_execution
# Restore the cache before cloning the repository because the clone script re-uses
# the restored repository if present. This reduces the amount of times the components
# repository needs to be cloned (this is slow and increases based on commits in the repo).
- restore_cache:
keys:
- *material_unit_tests_cache_key
- *material_unit_tests_cache_key_short
- run:
name: "Fetching Material repository"
command: ./scripts/ci/clone_angular_material_repo.sh
- run:
# Run yarn install to fetch the Bazel binaries as used in the Material repo.
name: Installing Material dependencies.
command: yarn --cwd ${MATERIAL_REPO_TMP_DIR} install --frozen-lockfile --non-interactive
- save_cache:
key: *material_unit_tests_cache_key
paths:
# Material directory must be kept in sync with the `$MATERIAL_REPO_TMP_DIR` env variable.
# It needs to be hardcoded here, because env variables interpolation is not supported.
- "/tmp/material2"
- run:
name: "Setup Bazel RBE remote execution in Material repo"
command: |
./.circleci/setup-rbe.sh "${MATERIAL_REPO_TMP_DIR}/.bazelrc.user"
- run:
name: "Running Material unit tests"
command: ./scripts/ci/run_angular_material_unit_tests.sh
test_zonejs:
<<: *job_defaults
steps:
- *attach_workspace
- *init_environment
# Install
- run: yarn --cwd packages/zone.js install --frozen-lockfile --non-interactive
# Run zone.js tools tests
- run: yarn --cwd packages/zone.js promisetest
- run: yarn --cwd packages/zone.js promisefinallytest
- run: yarn bazel build //packages/zone.js:npm_package &&
cp dist/bin/packages/zone.js/npm_package/dist/zone-mix.js ./packages/zone.js/test/extra/ &&
cp dist/bin/packages/zone.js/npm_package/dist/zone-patch-electron.js ./packages/zone.js/test/extra/ &&
yarn --cwd packages/zone.js electrontest
# Windows jobs
# Docs: https://circleci.com/docs/2.0/hello-world-windows/
# Skipping workspace for now because it fails to extract on windows.
# TODO: when CircleCI fixes it, split this single job into install/test ones.
# Notes:
# - windows needs its own cache key because binaries in node_modules are different.
# - windows might need its own workspace for the same reason.
test_win:
<<: *job_defaults_win
steps:
- checkout
- *init_environment_win
- *post_checkout_win
# TODO: windows cache restoration is currently failing. Re-enable when it's fixed.
# Example failure: https://circleci.com/gh/angular/angular/423738
# - *restore_cache_win
- *setup_circleci_bazel_config_win
- run: yarn install --frozen-lockfile --non-interactive
# Install @bazel/bazel globally and use that for the first run.
# Workaround for https://github.com/bazelbuild/rules_nodejs/issues/894
- run: yarn global add @bazel/bazel@$env:BAZEL_VERSION
- run: bazel info
# Ran into a command parsing problem where `-browser:chromium-local` was converted to
# `-browser: chromium-local` (a space was added) in https://circleci.com/gh/angular/angular/357511.
# Probably a powershell command parsing thing. This way there's no problem.
- run:
command: yarn circleci-win-ve
no_output_timeout: 45m
# - save_cache:
# key: *cache_key_win
# paths:
# - "node_modules"
# - "C:/Users/circleci/bazel_repository_cache"
test_ivy_aot_win:
<<: *job_defaults_win
steps:
- checkout
- *init_environment_win
- *post_checkout_win
# - *restore_cache_win
- *setup_circleci_bazel_config_win
- run: yarn install --frozen-lockfile --non-interactive
- run: yarn global add @bazel/bazel@$env:BAZEL_VERSION
- run: bazel info
- run:
command: yarn circleci-win-ivy
no_output_timeout: 45m
workflows:
version: 2
default_workflow:
jobs:
- setup:
filters:
branches:
ignore: g3
- lint:
requires:
- setup
- test:
requires:
- setup
- test_ivy_aot:
requires:
- setup
- build-npm-packages:
requires:
- setup
- build-ivy-npm-packages:
requires:
- setup
- legacy-misc-tests:
requires:
- build-npm-packages
- legacy-unit-tests-saucelabs:
requires:
- setup
- test_saucelabs_bazel:
requires:
- setup
# This job is currently a PoC and a subset of `legacy-unit-tests-saucelabs`. Running on
# master only to avoid wasting resources.
#
# TODO: Run this job on all branches (including PRs) as soon as it is not a PoC.
filters:
branches:
only: master
- test_aio:
requires:
- setup
- deploy_aio:
requires:
- test_aio
- test_aio_local:
requires:
- build-npm-packages
- test_aio_local_ivy:
requires:
- build-npm-packages
- test_aio_tools:
requires:
- build-npm-packages
- test_docs_examples:
requires:
- build-npm-packages
- test_docs_examples_ivy:
requires:
- build-ivy-npm-packages
- aio_preview:
requires:
- setup
# Only run on PR builds. (There can be no previews for non-PR builds.)
filters:
branches:
only: /pull\/\d+/
- test_aio_preview:
requires:
- aio_preview
- integration_test:
requires:
- build-npm-packages
- publish_snapshot:
# Note: no filters on this job because we want it to run for all upstream branches
# We'd really like to filter out pull requests here, but not yet available:
# https://discuss.circleci.com/t/workflows-pull-request-filter/14396/4
# Instead, the job just exits immediately at the first step.
requires:
# Only publish if tests and integration tests pass
- test
- test_ivy_aot
- integration_test
# Only publish if `aio`/`docs` tests using the locally built Angular packages pass
- test_aio_local
- test_aio_local_ivy
- test_docs_examples
- test_docs_examples_ivy
# Get the artifacts to publish from the build-packages-dist job
# since the publishing script expects the legacy outputs layout.
- build-npm-packages
- build-ivy-npm-packages
- legacy-unit-tests-saucelabs
- legacy-misc-tests
- material-unit-tests:
requires:
- build-ivy-npm-packages
- test_zonejs:
requires:
- setup
# Windows jobs very slow so we run it on non-PRs only for now.
# TODO: remove the filter when CircleCI makes Windows FS faster.
# The Windows jobs are only run after their non-windows counterparts finish successfully.
# This isn't strictly necessary as there is no artifact dependency, but helps economize
# CI resources by not attempting to build when we know should fail.
- test_win:
requires:
- test
filters:
branches:
ignore:
- /pull\/.*/
- test_ivy_aot_win:
requires:
- test_ivy_aot
filters:
branches:
ignore:
- /pull\/.*/
aio_monitoring:
jobs:
- setup
- aio_monitoring_stable:
requires:
- setup
- aio_monitoring_next:
requires:
- setup
triggers:
- schedule:
# Runs AIO monitoring jobs at 10:00AM every day.
cron: "0 10 * * *"
filters:
branches:
only:
- master