Skip to content

Move change detection to separate workflow in CI #122336

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
131 changes: 13 additions & 118 deletions .github/workflows/build.yml
Original file line number Diff line number Diff line change
@@ -1,8 +1,5 @@
name: Tests

# gh-84728: "paths-ignore" is not used to skip documentation-only PRs, because
# it prevents to mark a job as mandatory. A PR cannot be merged if a job is
# mandatory but not scheduled because of "paths-ignore".
on:
workflow_dispatch:
push:
Expand All @@ -23,121 +20,19 @@ concurrency:

jobs:
check_source:
name: 'Check for source changes'
runs-on: ubuntu-latest
timeout-minutes: 10
outputs:
# Some of the referenced steps set outputs conditionally and there may be
# cases when referencing them evaluates to empty strings. It is nice to
# work with proper booleans so they have to be evaluated through JSON
# conversion in the expressions. However, empty strings used like that
# may trigger all sorts of undefined and hard-to-debug behaviors in
# GitHub Actions CI/CD. To help with this, all of the outputs set here
# that are meant to be used as boolean flags (and not arbitrary strings),
# MUST have fallbacks with default values set. A common pattern would be
# to add ` || false` to all such expressions here, in the output
# definitions. They can then later be safely used through the following
# idiom in job conditionals and other expressions. Here's some examples:
#
# if: fromJSON(needs.check_source.outputs.run-docs)
#
# ${{
# fromJSON(needs.check_source.outputs.run_tests)
# && 'truthy-branch'
# || 'falsy-branch'
# }}
#
run-docs: ${{ steps.docs-changes.outputs.run-docs || false }}
run-win-msi: ${{ steps.win-msi-changes.outputs.run-win-msi || false }}
run_tests: ${{ steps.check.outputs.run_tests || false }}
run_hypothesis: ${{ steps.check.outputs.run_hypothesis || false }}
run_cifuzz: ${{ steps.check.outputs.run_cifuzz || false }}
config_hash: ${{ steps.config_hash.outputs.hash }} # str
steps:
- uses: actions/checkout@v4
- name: Check for source changes
id: check
run: |
if [ -z "$GITHUB_BASE_REF" ]; then
echo "run_tests=true" >> $GITHUB_OUTPUT
else
git fetch origin $GITHUB_BASE_REF --depth=1
# git diff "origin/$GITHUB_BASE_REF..." (3 dots) may be more
# reliable than git diff "origin/$GITHUB_BASE_REF.." (2 dots),
# but it requires to download more commits (this job uses
# "git fetch --depth=1").
#
# git diff "origin/$GITHUB_BASE_REF..." (3 dots) works with Git
# 2.26, but Git 2.28 is stricter and fails with "no merge base".
#
# git diff "origin/$GITHUB_BASE_REF.." (2 dots) should be enough on
# GitHub, since GitHub starts by merging origin/$GITHUB_BASE_REF
# into the PR branch anyway.
#
# https://github.com/python/core-workflow/issues/373
git diff --name-only origin/$GITHUB_BASE_REF.. | grep -qvE '(\.rst$|^Doc|^Misc|^\.pre-commit-config\.yaml$|\.ruff\.toml$|\.md$|mypy\.ini$)' && echo "run_tests=true" >> $GITHUB_OUTPUT || true
fi

# Check if we should run hypothesis tests
GIT_BRANCH=${GITHUB_BASE_REF:-${GITHUB_REF#refs/heads/}}
echo $GIT_BRANCH
if $(echo "$GIT_BRANCH" | grep -q -w '3\.\(8\|9\|10\|11\)'); then
echo "Branch too old for hypothesis tests"
echo "run_hypothesis=false" >> $GITHUB_OUTPUT
else
echo "Run hypothesis tests"
echo "run_hypothesis=true" >> $GITHUB_OUTPUT
fi

# oss-fuzz maintains a configuration for fuzzing the main branch of
# CPython, so CIFuzz should be run only for code that is likely to be
# merged into the main branch; compatibility with older branches may
# be broken.
FUZZ_RELEVANT_FILES='(\.c$|\.h$|\.cpp$|^configure$|^\.github/workflows/build\.yml$|^Modules/_xxtestfuzz)'
if [ "$GITHUB_BASE_REF" = "main" ] && [ "$(git diff --name-only origin/$GITHUB_BASE_REF.. | grep -qE $FUZZ_RELEVANT_FILES; echo $?)" -eq 0 ]; then
# The tests are pretty slow so they are executed only for PRs
# changing relevant files.
echo "Run CIFuzz tests"
echo "run_cifuzz=true" >> $GITHUB_OUTPUT
else
echo "Branch too old for CIFuzz tests; or no C files were changed"
echo "run_cifuzz=false" >> $GITHUB_OUTPUT
fi
- name: Compute hash for config cache key
id: config_hash
run: |
echo "hash=${{ hashFiles('configure', 'configure.ac', '.github/workflows/build.yml') }}" >> $GITHUB_OUTPUT
- name: Get a list of the changed documentation-related files
if: github.event_name == 'pull_request'
id: changed-docs-files
uses: Ana06/get-changed-files@v2.3.0
with:
filter: |
Doc/**
Misc/**
.github/workflows/reusable-docs.yml
format: csv # works for paths with spaces
- name: Check for docs changes
if: >-
github.event_name == 'pull_request'
&& steps.changed-docs-files.outputs.added_modified_renamed != ''
id: docs-changes
run: |
echo "run-docs=true" >> "${GITHUB_OUTPUT}"
- name: Get a list of the MSI installer-related files
id: changed-win-msi-files
uses: Ana06/get-changed-files@v2.3.0
with:
filter: |
Tools/msi/**
.github/workflows/reusable-windows-msi.yml
format: csv # works for paths with spaces
- name: Check for changes in MSI installer-related files
if: >-
steps.changed-win-msi-files.outputs.added_modified_renamed != ''
id: win-msi-changes
run: |
echo "run-win-msi=true" >> "${GITHUB_OUTPUT}"
name: Change detection
# To use boolean outputs from this job, parse them as JSON.
# Here's some examples:
#
# if: fromJSON(needs.check_source.outputs.run-docs)
#
# ${{
# fromJSON(needs.check_source.outputs.run_tests)
# && 'truthy-branch'
# || 'falsy-branch'
# }}
#
uses: ./.github/workflows/reusable-change-detection.yml

check-docs:
name: Docs
Expand Down
150 changes: 150 additions & 0 deletions .github/workflows/reusable-change-detection.yml
Original file line number Diff line number Diff line change
@@ -0,0 +1,150 @@
---

name: Change detection

on: # yamllint disable-line rule:truthy
workflow_call:
outputs:
# Some of the referenced steps set outputs conditionally and there may be
# cases when referencing them evaluates to empty strings. It is nice to
# work with proper booleans so they have to be evaluated through JSON
# conversion in the expressions. However, empty strings used like that
# may trigger all sorts of undefined and hard-to-debug behaviors in
# GitHub Actions CI/CD. To help with this, all of the outputs set here
# that are meant to be used as boolean flags (and not arbitrary strings),
# MUST have fallbacks with default values set. A common pattern would be
# to add ` || false` to all such expressions here, in the output
# definitions. They can then later be safely used through the following
# idiom in job conditionals and other expressions. Here's some examples:
#
# if: fromJSON(needs.change-detection.outputs.run-docs)
#
# ${{
# fromJSON(needs.change-detection.outputs.run-tests)
# && 'truthy-branch'
# || 'falsy-branch'
# }}
#
config_hash:
description: Config hash value for use in cache keys
value: ${{ jobs.compute-changes.outputs.config-hash }} # str
run-docs:
description: Whether to build the docs
value: ${{ jobs.compute-changes.outputs.run-docs || false }} # bool
run_tests:
description: Whether to run the regular tests
value: ${{ jobs.compute-changes.outputs.run-tests || false }} # bool
run-win-msi:
description: Whether to run the MSI installer smoke tests
value: >- # bool
${{ jobs.compute-changes.outputs.run-win-msi || false }}
run_hypothesis:
description: Whether to run the Hypothesis tests
value: >- # bool
${{ jobs.compute-changes.outputs.run-hypothesis || false }}
run_cifuzz:
description: Whether to run the CIFuzz job
value: >- # bool
${{ jobs.compute-changes.outputs.run-cifuzz || false }}

jobs:
compute-changes:
name: Compute changed files
runs-on: ubuntu-latest
timeout-minutes: 10
outputs:
config-hash: ${{ steps.config-hash.outputs.hash }}
run-cifuzz: ${{ steps.check.outputs.run-cifuzz }}
run-docs: ${{ steps.docs-changes.outputs.run-docs }}
run-hypothesis: ${{ steps.check.outputs.run-hypothesis }}
run-tests: ${{ steps.check.outputs.run-tests }}
run-win-msi: ${{ steps.win-msi-changes.outputs.run-win-msi }}
steps:
- run: >-
echo '${{ github.event_name }}'
- uses: actions/checkout@v4
- name: Check for source changes
id: check
run: |
if [ -z "$GITHUB_BASE_REF" ]; then
echo "run-tests=true" >> $GITHUB_OUTPUT
else
git fetch origin $GITHUB_BASE_REF --depth=1
# git diff "origin/$GITHUB_BASE_REF..." (3 dots) may be more
# reliable than git diff "origin/$GITHUB_BASE_REF.." (2 dots),
# but it requires to download more commits (this job uses
# "git fetch --depth=1").
#
# git diff "origin/$GITHUB_BASE_REF..." (3 dots) works with Git
# 2.26, but Git 2.28 is stricter and fails with "no merge base".
#
# git diff "origin/$GITHUB_BASE_REF.." (2 dots) should be enough on
# GitHub, since GitHub starts by merging origin/$GITHUB_BASE_REF
# into the PR branch anyway.
#
# https://github.com/python/core-workflow/issues/373
git diff --name-only origin/$GITHUB_BASE_REF.. | grep -qvE '(\.rst$|^Doc|^Misc|^\.pre-commit-config\.yaml$|\.ruff\.toml$|\.md$|mypy\.ini$)' && echo "run-tests=true" >> $GITHUB_OUTPUT || true
fi

# Check if we should run hypothesis tests
GIT_BRANCH=${GITHUB_BASE_REF:-${GITHUB_REF#refs/heads/}}
echo $GIT_BRANCH
if $(echo "$GIT_BRANCH" | grep -q -w '3\.\(8\|9\|10\|11\)'); then
echo "Branch too old for hypothesis tests"
echo "run-hypothesis=false" >> $GITHUB_OUTPUT
else
echo "Run hypothesis tests"
echo "run-hypothesis=true" >> $GITHUB_OUTPUT
fi

# oss-fuzz maintains a configuration for fuzzing the main branch of
# CPython, so CIFuzz should be run only for code that is likely to be
# merged into the main branch; compatibility with older branches may
# be broken.
FUZZ_RELEVANT_FILES='(\.c$|\.h$|\.cpp$|^configure$|^\.github/workflows/build\.yml$|^Modules/_xxtestfuzz)'
if [ "$GITHUB_BASE_REF" = "main" ] && [ "$(git diff --name-only origin/$GITHUB_BASE_REF.. | grep -qE $FUZZ_RELEVANT_FILES; echo $?)" -eq 0 ]; then
# The tests are pretty slow so they are executed only for PRs
# changing relevant files.
echo "Run CIFuzz tests"
echo "run-cifuzz=true" >> $GITHUB_OUTPUT
else
echo "Branch too old for CIFuzz tests; or no C files were changed"
echo "run-cifuzz=false" >> $GITHUB_OUTPUT
fi
- name: Compute hash for config cache key
id: config-hash
run: |
echo "hash=${{ hashFiles('configure', 'configure.ac', '.github/workflows/build.yml') }}" >> $GITHUB_OUTPUT
- name: Get a list of the changed documentation-related files
if: github.event_name == 'pull_request'
id: changed-docs-files
uses: Ana06/get-changed-files@v2.3.0
with:
filter: |
Doc/**
Misc/**
.github/workflows/reusable-docs.yml
format: csv # works for paths with spaces
- name: Check for docs changes
if: >-
github.event_name == 'pull_request'
&& steps.changed-docs-files.outputs.added_modified_renamed != ''
id: docs-changes
run: |
echo "run-docs=true" >> "${GITHUB_OUTPUT}"
- name: Get a list of the MSI installer-related files
id: changed-win-msi-files
uses: Ana06/get-changed-files@v2.3.0
with:
filter: |
Tools/msi/**
.github/workflows/reusable-windows-msi.yml
format: csv # works for paths with spaces
- name: Check for changes in MSI installer-related files
if: >-
steps.changed-win-msi-files.outputs.added_modified_renamed != ''
id: win-msi-changes
run: |
echo "run-win-msi=true" >> "${GITHUB_OUTPUT}"

...
Loading