Files
carbon-lang/.github/workflows/tests.yaml
T
Chandler Carruthandjosh11b 49d46bd8a4 Add up to 10 retries to our test action. (#3401)
This captures the exit code of Bazel and checks for success or permanent
errors on each attempt. It also sleeps a small amount between attempts.
We should be able to increase the retries and sleeps as needed to
minimize flakiness here, and Bazel should even persist incremental
progress efficiently. Hopefully this helps reduce the failure rate of
our CI.

It also changes how we build on a `push` to use a single Bazel clause to
hold this logic.

Managed to get one of the download failures when testing this, and the
retry logic worked but there was a bug in the success logic.

Otherwise seems to work:
- Synthetic failure:
https://github.com/carbon-language/carbon-lang/actions/runs/6872431707/job/18690894069
- Success:
https://github.com/carbon-language/carbon-lang/actions/runs/6872461061

---------

Co-authored-by: josh11b <josh11b@users.noreply.github.com>
2023-11-15 15:34:50 +00:00

307 lines
12 KiB
YAML

# Part of the Carbon Language project, under the Apache License v2.0 with LLVM
# Exceptions. See /LICENSE for license information.
# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
name: test
on:
push:
branches: [trunk]
pull_request:
merge_group:
# Cancel previous workflows on the PR when there are multiple fast commits.
# https://docs.github.com/en/actions/using-workflows/workflow-syntax-for-github-actions#concurrency
concurrency:
group: ${{ github.workflow }}-${{ github.head_ref || github.ref }}
cancel-in-progress: true
jobs:
test:
strategy:
matrix:
# At present, these images are newer than "latest". We use them to test
# against more recent tooling versions.
# https://github.com/actions/runner-images
os: [ubuntu-22.04, macos-12]
build_mode: [fastbuild, opt]
runs-on: ${{ matrix.os }}
steps:
# Ubuntu images start with 23GB available, and this adds 14GB more. For
# comparison, MacOS images have >100GB free.
- name: Free up disk space (Ubuntu)
if: matrix.os == 'ubuntu-22.04'
uses: jlumbroso/free-disk-space@v1.2.0
with:
android: true
dotnet: true
haskell: true
# Although we could delete more, if we run into a limit, it provides a
# little flexibility to get space while trying to shrink the build.
# There's also support for docker images at head (1.2.0 is still
# the latest release).
large-packages: false
swap-storage: false
# Checkout the pull request head or the branch.
- name: Checkout pull request
if: github.event_name == 'pull_request'
uses: actions/checkout@v3
with:
ref: ${{ github.event.pull_request.head.sha }}
- name: Checkout branch
if: github.event_name != 'pull_request'
uses: actions/checkout@v3
# Tests should only run on applicable paths, but we still need to have an
# action run for the merge queue. We filter steps based on the paths here,
# and condition steps on the output.
- id: filter
uses: dorny/paths-filter@v2
with:
filters: |
ignore:
- '**/*.md'
- 'LICENSE'
- 'CODEOWNERS'
- '.git*'
# Setup Python and related tools.
- uses: actions/setup-python@v4
if: steps.filter.outputs.ignore == 'false'
with:
# Match the min version listed in docs/project/contribution_tools.md
python-version: '3.9'
# Use LLVM following:
# https://github.com/actions/runner-images/blob/main/images/macos/macos-12-Readme.md
# Both 14 and 15 are candidates because GitHub is testing new images.
- name: Setup LLVM and Clang (macOS)
if: steps.filter.outputs.ignore == 'false' && matrix.os == 'macos-12'
run: |
LLVM_PATH="$(brew --prefix llvm@15)"
if [[ ! -e "${LLVM_PATH}" ]]; then
LLVM_PATH="$(brew --prefix llvm@14)"
fi
echo "Using ${LLVM_PATH}"
echo "${LLVM_PATH}/bin" >> $GITHUB_PATH
echo '*** ls "${LLVM_PATH}"'
ls "${LLVM_PATH}"
echo '*** ls "${LLVM_PATH}/bin"'
ls "${LLVM_PATH}/bin"
# Use LLVM following:
# https://github.com/actions/runner-images/blob/main/images/linux/Ubuntu2204-Readme.md
# Both 14 and 15 are candidates for forwards compatibility, although 15
# isn't provided.
- name: Setup LLVM and Clang (Ubuntu)
if:
steps.filter.outputs.ignore == 'false' && matrix.os == 'ubuntu-22.04'
run: |
# TODO: Re-enable once llvm-15 is working.
# https://github.com/actions/runner-images/issues/8253
# LLVM_PATH="/usr/lib/llvm-15"
# if [[ ! -e "${LLVM_PATH}" ]]; then
LLVM_PATH="/usr/lib/llvm-14"
# fi
echo "Using ${LLVM_PATH}"
echo "${LLVM_PATH}/bin" >> $GITHUB_PATH
echo '*** ls "${LLVM_PATH}"'
ls "${LLVM_PATH}"
echo '*** ls "${LLVM_PATH}/bin"'
ls "${LLVM_PATH}/bin"
# Print the various tool paths and versions to help in debugging.
- name: Print tool debugging info
if: steps.filter.outputs.ignore == 'false'
run: |
echo '*** PATH'
echo $PATH
echo '*** bazelisk'
which bazelisk
bazelisk --version
echo '*** python'
which python
python --version
echo '*** clang'
which clang
clang --version
echo '*** clang++'
which clang++
clang++ --version
# Disable uploads when the remote cache is read-only.
- name: Set up remote cache access (read-only)
if:
steps.filter.outputs.ignore == 'false' && github.event_name ==
'pull_request'
run: |
echo "remote_cache_upload=--remote_upload_local_results=false" \
>> $GITHUB_ENV
# Provide a cache key when the remote cache is read-write.
- name: Set up remote cache access (read-write)
if:
steps.filter.outputs.ignore == 'false' && github.event_name !=
'pull_request'
env:
REMOTE_CACHE_KEY: ${{ secrets.CARBON_BUILDS_GITHUB }}
run: |
echo "$REMOTE_CACHE_KEY" | base64 -d > $HOME/remote_cache_key.json
echo "remote_cache_upload=--google_credentials=$HOME/remote_cache_key.json" \
>> $GITHUB_ENV
# We need to replace the `.` with a `_` for the build cache.
- name: Setup LLVM and Clang (macOS)
if: steps.filter.outputs.ignore == 'false' && matrix.os == 'macos-12'
run: |
echo "os_for_cache=macos-12" >> $GITHUB_ENV
- name: Setup LLVM and Clang (Ubuntu)
if:
steps.filter.outputs.ignore == 'false' && matrix.os == 'ubuntu-22.04'
run: |
echo "os_for_cache=ubuntu-22_04" >> $GITHUB_ENV
# Add our bazel configuration and print basic info to ease debugging.
- name: Configure Bazel and print info
if: steps.filter.outputs.ignore == 'false'
env:
# Add a cache version for changes that bazel won't otherwise detect,
# like llvm version changes.
CACHE_VERSION: 1
run: |
cat >user.bazelrc <<EOF
# Enable remote cache for our CI but minimize downloads.
build --remote_cache=https://storage.googleapis.com/carbon-builds-github-v${CACHE_VERSION}-${{ env.os_for_cache }}
build --remote_download_minimal
build ${{ env.remote_cache_upload }}
# Set an artificially high jobs count. This flag controls the number
# of concurrency Bazel itself uses, which is essential for actions
# that are internally blocked on for example downloading results form
# the cache above. Without setting this high, Bazel will pick a small
# number based on the available host CPUs and the reality will be a
# long chain of largely serialized download events with little or no
# usage of the host machine. Fortunately, local actions are
# *separately* gated on '--local_*_resources' that will avoid a large
# jobs value overwhelming the host. There is a bug to make downloads
# behave completely asynchronously and remove the need for this filed
# back in 2018 but work seemed to not finish:
# https://github.com/bazelbuild/bazel/issues/6394
#
# There is a new effort (yay!) but until then it seems worth using the
# workaround of a high jobs value. The biggest downside (increased
# heap usage) seems like it isn't currently a big loss for our builds.
#
# Higher values like 50 have led to CI failures with network errors
# and IOExceptions, see
# https://discord.com/channels/655572317891461132/707150492370862090/1151605725576056934
build --jobs=32
# General build options.
build --verbose_failures
test --test_output=errors
EOF
bazelisk info
# Just for visibility, print space before and after the build.
- name: Disk space before build
if: steps.filter.outputs.ignore == 'false'
run: df -h
# Build and run all targets on branch pushes to ensure we always have a
# clean tree. We don't expect this to be an interactive path and so don't
# optimize the latency of this step.
- name: Compute impacted pull request targets (for push)
if:
steps.filter.outputs.ignore == 'false' && github.event_name == 'push'
env:
TARGETS_FILE: ${{ runner.temp }}/targets
run: |
echo "//..." >$TARGETS_FILE
# Compute the set of possible rules impacted by this change using
# Bazel-based diffing. This lets PRs and the merge queue have a much more
# efficient test CI action by avoiding even enumerating (and downloading)
# all of the unaffected Bazel targets.
- name: Compute impacted pull request targets
if:
steps.filter.outputs.ignore == 'false' && github.event_name != 'push'
env:
# Compute the base SHA from the different event structures.
GIT_BASE_SHA:
${{ github.event_name == 'pull_request' &&
github.event.pull_request.base.sha ||
github.event.merge_group.base_sha }}
TARGETS_FILE: ${{ runner.temp }}/targets
run: |
# First fetch the relevant base into the git repository.
git fetch --depth=1 origin $GIT_BASE_SHA
# Then use `target-determinator` as wrapped by our script.
./scripts/target_determinator.py $GIT_BASE_SHA >$TARGETS_FILE
# Build and run just the tests impacted by the PR or merge group.
- name: Test (${{ matrix.build_mode }})
if: steps.filter.outputs.ignore == 'false'
env:
# 'libtool_check_unique failed to generate' workaround.
# https://github.com/bazelbuild/bazel/issues/14113#issuecomment-999794586
BAZEL_USE_CPP_ONLY_TOOLCHAIN: 1
TARGETS_FILE: ${{ runner.temp }}/targets
run: |
# Bazel requires a test target to run the test command. There may be
# no targets or there may only be non-test targets that we want to
# build, so simply inject an explicit no-op test target.
echo "//scripts:no_op_test" >> $TARGETS_FILE
for i in {1..5}; do
if (( $i == 4 )); then
# Decrease the jobs sharply if we see repeated failures to try to
# work around transient network errors even if it makes things
# slower.
echo "build --jobs=4" >>user.bazelrc
fi
bazel_exit=0
bazelisk test -c ${{ matrix.build_mode }} \
--target_pattern_file=$TARGETS_FILE || bazel_exit=$?
# If we succeed, we're done.
if (( $bazel_exit == 0 )); then
break
fi
# Several error codes are reliably permanent, break immediately.
# `1` -- The build failed.
# `2` -- Command line or environment problem.
# `3` -- Tests failed or timed out, we don't retry at this layer
# on execution timeout.
# `4` -- No tests found, which should be impossible here.
# `8` -- Explicitly interrupted build.
#
# Note that `36` is documented as "likely permanent", but we retry
# it as most of our transient failures actually produce that error
# code.
if (( $bazel_exit == 1 || $bazel_exit == 2 || $bazel_exit == 3 || \
$bazel_exit == 4 || $bazel_exit == 8 || $bazel_exit == 8 ))
then
break
fi
echo "Retrying a failed build as it may be transient..."
# Also sleep a bit to try to skip over transient machine load.
sleep $i
done
# Propagate the Bazel exit code.
exit $bazel_exit
# See "Disk space before build".
- name: Disk space after build
if: steps.filter.outputs.ignore == 'false'
run: df -h