ci(pr-review-poster): add line-anchored review poster and migrate clang-tidy (#27028)

* ci(pr-review-poster): add line-anchored review poster and migrate clang-tidy

Adds a generic PR review-comment poster as a sibling of the issue-comment
poster from #27021. Replaces platisd/clang-tidy-pr-comments@v1 in the
Static Analysis workflow with an in-tree, fork-friendly producer + poster
pair so fork PRs get inline clang-tidy annotations on the Files changed
tab without trusting a third-party action with a write token.

Architecture mirrors pr-comment-poster: a producer (clang-tidy.yml) runs
inside the px4-dev container and writes a `pr-review` artifact containing
manifest.json and a baked comments.json. A separate workflow_run-triggered
poster runs on ubuntu-latest with the base-repo write token, validates the
artifact, dismisses any stale matching review, and posts a fresh review
on the target PR. The poster never checks out PR code and only ever reads
two opaque JSON files from the artifact.

Stale-review dismissal is restricted to reviews authored by
github-actions[bot] AND whose body contains the producer's marker. A fork
cannot impersonate the bot login or inject the marker into a human
reviewer's body, so the poster can never dismiss a human review. APPROVE
events are explicitly forbidden so a bot cannot approve a pull request.

To avoid duplicating ~120 lines of HTTP plumbing between the two posters,
the GitHub REST helpers (single-request, pagination, error handling) are
extracted into Tools/ci/_github_helpers.py with a small GitHubClient
class. The existing pr-comment-poster.py is refactored to use it; net
change is roughly -80 lines on that script. The shared module is
sparse-checked-out alongside each poster script and is stdlib only.

The clang-tidy producer reuses MIT-licensed translation logic from
platisd/clang-tidy-pr-comments (generate_review_comments,
reorder_diagnostics, get_diff_line_ranges_per_file and helpers) under a
preserved attribution header. The HTTP layer is rewritten on top of
_github_helpers so the producer does not pull in `requests`. Conversation
resolution (the GraphQL path) is intentionally dropped for v1.

clang-tidy.yml now produces the pr-review artifact in the same job as
the build, so the cross-runner compile_commands.json hand-off and
workspace-path rewriting are no longer needed and the
post_clang_tidy_comments job is removed.

Signed-off-by: Ramon Roche <mrpollo@gmail.com>

* ci(workflows): bump action versions to clear Node 20 deprecation

GitHub has deprecated the Node 20 runtime for Actions as of
September 16, 2026. Bump the pinned action versions in the three poster
workflows to the latest majors, all of which run on Node 24:

  actions/checkout         v4 -> v6
  actions/github-script    v7 -> v8
  actions/upload-artifact  v4 -> v7

No behavior changes on our side: upload-artifact v5/v6/v7 only added an
optional direct-file-upload mode we do not use, and checkout v5/v6 are
runtime-only bumps. The security-invariant comment headers in both
poster workflows are updated to reference the new version so they stay
accurate.

Signed-off-by: Ramon Roche <mrpollo@gmail.com>

* ci(pr-posters): skip job when producer was not a pull_request event

Both poster workflows previously ran on every workflow_run completion of
their listed producers and then silently no-oped inside the script when
the triggering producer run was a push-to-main (or any other non-PR
event). That made the UI ambiguous: the job was always green, never
showed the reason it did nothing, and looked like a failure whenever
someone clicked in looking for the comment that was never there.

Gate the job at the workflow level on
github.event.workflow_run.event == 'pull_request'. Non-PR producer runs
now surface as a clean "Skipped" entry in the run list, which is
self-explanatory and needs no in-script summary plumbing.

Signed-off-by: Ramon Roche <mrpollo@gmail.com>

---------

Signed-off-by: Ramon Roche <mrpollo@gmail.com>
This commit is contained in:
Ramon Roche 2026-04-09 11:54:29 -06:00 committed by GitHub
parent e8c19a2006
commit 9e93fd753e
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
7 changed files with 1424 additions and 222 deletions

View File

@ -21,11 +21,12 @@ jobs:
runs-on: [runs-on, runner=16cpu-linux-x64, "run-id=${{ github.run_id }}", "extras=s3-cache"]
container:
image: ghcr.io/px4/px4-dev:v1.17.0-rc2
outputs:
has_findings: ${{ steps.clang_tidy.outputs.has_findings }}
permissions:
contents: read
pull-requests: read
steps:
- uses: runs-on/action@v2
- uses: actions/checkout@v4
- uses: actions/checkout@v6
with:
fetch-depth: 0
fetch-tags: true
@ -47,96 +48,50 @@ jobs:
run: |
if [ "${{ github.event_name }}" != "pull_request" ]; then
make -j$(nproc) clang-tidy
echo "has_findings=false" >> $GITHUB_OUTPUT
else
output=$(python3 Tools/ci/run-clang-tidy-pr.py origin/${{ github.base_ref }} 2>&1)
exit_code=$?
echo "$output"
# Helper prints this message on both early-exit paths
# (no changed C++ files, or no matching TUs in the compile DB)
if echo "$output" | grep -q "skipping clang-tidy"; then
echo "has_findings=false" >> $GITHUB_OUTPUT
else
echo "has_findings=true" >> $GITHUB_OUTPUT
fi
exit $exit_code
python3 Tools/ci/run-clang-tidy-pr.py origin/${{ github.base_ref }}
fi
- name: Upload compile_commands.json
# On PRs, also produce a `pr-review` artifact for the PR Review Poster
# workflow to consume. clang-tidy-diff-18 emits a unified fixes.yml that
# the producer script translates into line-anchored review comments.
# Running this inside the same container as the build means there is no
# workspace-path rewriting and no cross-runner artifact handoff.
- name: Export clang-tidy fixes for PR review
if: always() && github.event_name == 'pull_request'
uses: actions/upload-artifact@v4
run: |
mkdir -p pr-review
git diff -U0 origin/${{ github.base_ref }}...HEAD \
| clang-tidy-diff-18.py -p1 \
-path build/px4_sitl_default-clang \
-export-fixes pr-review/fixes.yml \
-j0 || true
- name: Build pr-review artifact
if: always() && github.event_name == 'pull_request'
env:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
run: |
python3 Tools/ci/clang-tidy-fixes-to-review.py \
--fixes pr-review/fixes.yml \
--repo-root "$GITHUB_WORKSPACE" \
--repo "$GITHUB_REPOSITORY" \
--pr-number "${{ github.event.pull_request.number }}" \
--commit-sha "${{ github.event.pull_request.head.sha }}" \
--out-dir pr-review \
--event REQUEST_CHANGES
- name: Upload pr-review artifact
if: always() && github.event_name == 'pull_request'
uses: actions/upload-artifact@v7
with:
name: compile-commands
path: build/px4_sitl_default-clang/compile_commands.json
name: pr-review
path: |
pr-review/manifest.json
pr-review/comments.json
retention-days: 1
- uses: ./.github/actions/save-ccache
if: always()
with:
cache-primary-key: ${{ steps.ccache.outputs.cache-primary-key }}
post_clang_tidy_comments:
name: Clang-Tidy PR Annotations
needs: [clang_tidy]
runs-on: ubuntu-latest
permissions:
pull-requests: write
contents: read
if: >-
always()
&& github.event.pull_request
&& github.event.pull_request.head.repo.full_name == github.repository
&& (needs.clang_tidy.result == 'success' || needs.clang_tidy.result == 'failure')
&& needs.clang_tidy.outputs.has_findings == 'true'
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Install clang-tools (for clang-tidy-diff)
run: sudo apt-get install -y clang-tools
- name: Download compile_commands.json
uses: actions/download-artifact@v4
with:
name: compile-commands
path: build/px4_sitl_default-clang
- name: Run clang-tidy-diff and export fixes
run: |
# WHY WE REWRITE compile_commands.json PATHS
#
# The clang_tidy job runs on a RunsOn/AWS runner where the workspace
# root is "/__w/PX4-Autopilot/PX4-Autopilot". All absolute paths baked
# into compile_commands.json (the "file" and "directory" fields, and
# every -I include path in "command") use that prefix.
#
# This annotation job runs on a GitHub-hosted runner where the
# workspace root is "/home/runner/work/PX4-Autopilot/PX4-Autopilot".
# When clang-tidy-diff invokes clang-tidy, it reads the "directory"
# field from compile_commands.json and calls chdir() on it. Since
# the AWS-style path does not exist here, clang-tidy crashes with:
# LLVM ERROR: Cannot chdir into "/__w/.../build/px4_sitl_default-clang"
# and silently produces an empty fixes.yml, so no annotations are posted.
#
# Fix: rewrite all occurrences of the AWS workspace prefix to the
# current runner workspace ($GITHUB_WORKSPACE) before invoking
# clang-tidy-diff. Safe because compile_commands.json is a local
# scratch file pulled from the artifact; no source file is modified.
sed -i "s|/__w/PX4-Autopilot/PX4-Autopilot|${GITHUB_WORKSPACE}|g" \
build/px4_sitl_default-clang/compile_commands.json
mkdir -p clang-tidy-result
git diff -U0 origin/${{ github.base_ref }}...HEAD \
| clang-tidy-diff-18.py -p1 \
-path build/px4_sitl_default-clang \
-export-fixes clang-tidy-result/fixes.yml \
-j0 || true
- name: Annotate PR with clang-tidy findings
uses: platisd/clang-tidy-pr-comments@v1
with:
github_token: ${{ secrets.GITHUB_TOKEN }}
clang_tidy_fixes: clang-tidy-result/fixes.yml
request_changes: true
suggestions_per_comment: 10

View File

@ -40,12 +40,13 @@ name: PR Comment Poster
# manifest and body itself, NOT copied fork-controlled content into
# them. Producer workflows are responsible for that.
#
# 6. `actions/checkout@v4` below uses NO ref (so it pulls the base branch,
# 6. `actions/checkout@v6` below uses NO ref (so it pulls the base branch,
# the default-branch commit this workflow file was loaded from) AND uses
# sparse-checkout to materialize ONLY Tools/ci/pr-comment-poster.py. The
# rest of the repo never touches the workspace. This is safe: the only
# file the job executes is a base-repo Python script that was reviewed
# through normal code review, never anything from the PR.
# sparse-checkout to materialize ONLY Tools/ci/pr-comment-poster.py and
# its stdlib-only helper module Tools/ci/_github_helpers.py. The rest of
# the repo never touches the workspace. This is safe: the only files the
# job executes are base-repo Python scripts that were reviewed through
# normal code review, never anything from the PR.
#
# ==============================================================================
# ARTIFACT CONTRACT
@ -91,22 +92,29 @@ jobs:
post:
name: Post PR Comment
runs-on: ubuntu-latest
if: github.event.workflow_run.conclusion != 'cancelled'
# Only run for pull_request producer runs. Push-to-main and other
# non-PR triggers would have no comment to post, and silently no-oping
# inside the script made it look like the poster was broken. Gating at
# the job level surfaces those as a clean "Skipped" in the UI instead.
if: >-
github.event.workflow_run.conclusion != 'cancelled'
&& github.event.workflow_run.event == 'pull_request'
steps:
# Checkout runs first so the poster script is available AND so that
# actions/checkout@v4's default clean step does not delete the artifact
# actions/checkout@v6's default clean step does not delete the artifact
# zip that the next step writes into the workspace. Sparse-checkout
# restricts the materialized tree to just the poster script.
- name: Checkout poster script only
uses: actions/checkout@v4
uses: actions/checkout@v6
with:
sparse-checkout: |
Tools/ci/pr-comment-poster.py
Tools/ci/_github_helpers.py
sparse-checkout-cone-mode: false
- name: Download pr-comment artifact
id: download
uses: actions/github-script@v7
uses: actions/github-script@v8
with:
script: |
const artifacts = await github.rest.actions.listWorkflowRunArtifacts({

177
.github/workflows/pr-review-poster.yml vendored Normal file
View File

@ -0,0 +1,177 @@
name: PR Review Poster
# Generic PR review-comment poster. Sibling of "PR Comment Poster": that
# workflow posts sticky issue-style comments, this one posts line-anchored
# review comments on the "Files changed" tab. Any analysis workflow that
# wants to flag specific lines can produce a `pr-review` artifact and this
# workflow will dismiss any stale matching review and post a fresh one.
# Designed so analysis jobs running on untrusted fork PRs can still get
# their inline annotations posted back to the PR.
#
# ==============================================================================
# SECURITY INVARIANTS
# ==============================================================================
# This workflow runs on `workflow_run` which means it runs in the BASE REPO
# context with a WRITE token, even when the triggering PR comes from a fork.
# That is the entire reason it exists, and also the reason it is a loaded
# footgun. Anyone modifying this file MUST preserve the following invariants:
#
# 1. NEVER check out PR code. No `actions/checkout` with a ref. No git clone
# of a fork branch. No execution of scripts from the downloaded artifact.
# The ONLY things read from the artifact are `manifest.json` and
# `comments.json`, and both are treated as opaque data (JSON parsed by
# the poster script and the comment fields posted via the GitHub API).
#
# 2. `pr_number` is validated to be a positive integer before use.
# `marker` is validated to be printable ASCII only before use.
# `commit_sha` is validated to be 40 lowercase hex characters.
# `event` is validated against an allowlist of `COMMENT` and
# `REQUEST_CHANGES`. `APPROVE` is intentionally forbidden so a bot
# cannot approve a pull request. Validation happens inside
# Tools/ci/pr-review-poster.py which is checked out from the base
# branch, not from the artifact.
#
# 3. Comment bodies and the optional summary are passed to the GitHub API
# as JSON fields, never interpolated into a shell command string.
#
# 4. This workflow file lives on the default branch. `workflow_run` only
# loads workflow files from the default branch, so a fork cannot modify
# THIS file as part of a PR. The fork CAN cause this workflow to fire
# by triggering a producer workflow that uploads a `pr-review`
# artifact. That is intended.
#
# 5. The artifact-name filter (`pr-review`) is the only gate on which
# workflow runs get processed. Any workflow in this repo that uploads
# an artifact named `pr-review` is trusted to have written the
# manifest and comments itself, NOT copied fork-controlled content
# into them. Producer workflows are responsible for that.
#
# 6. `actions/checkout@v6` below uses NO ref (so it pulls the base branch,
# the default-branch commit this workflow file was loaded from) AND
# uses sparse-checkout to materialize ONLY
# Tools/ci/pr-review-poster.py and its stdlib-only helper module
# Tools/ci/_github_helpers.py. The rest of the repo never touches the
# workspace. This is safe: the only files the job executes are
# base-repo Python scripts that were reviewed through normal code
# review, never anything from the PR.
#
# 7. Stale-review dismissal is restricted to reviews whose AUTHOR is
# `github-actions[bot]` AND whose body contains the producer's
# marker. A fork PR cannot impersonate the bot login, and cannot
# inject the marker into a human reviewer's body without API
# access. Both filters together prevent the poster from ever
# dismissing a human review.
#
# ==============================================================================
# ARTIFACT CONTRACT
# ==============================================================================
# Producers upload an artifact named exactly `pr-review` containing:
#
# manifest.json:
# {
# "pr_number": 12345, // required, int > 0
# "marker": "<!-- pr-review-poster:clang-tidy -->", // required, printable ASCII
# "event": "REQUEST_CHANGES", // required, "COMMENT" | "REQUEST_CHANGES"
# "commit_sha": "0123456789abcdef0123456789abcdef01234567", // required, 40 hex chars
# "summary": "Optional review summary text" // optional
# }
#
# comments.json: JSON array of line-anchored review comment objects:
# [
# {"path": "src/foo.cpp", "line": 42, "side": "RIGHT", "body": "..."},
# {"path": "src/bar.hpp", "start_line": 10, "line": 15,
# "side": "RIGHT", "start_side": "RIGHT", "body": "..."}
# ]
#
# The `marker` string is used to find an existing matching review to
# dismiss before posting a new one. It MUST be unique per producer (e.g.
# include the producer name).
#
# Producers MUST write `pr_number` and `commit_sha` from their own
# workflow context (`github.event.pull_request.number` and
# `github.event.pull_request.head.sha`) and MUST NOT read either from any
# fork-controlled source.
on:
workflow_run:
# Producers that may upload a `pr-review` artifact. When a new
# producer is wired up, add its workflow name here. Runs of workflows
# not in this list will never trigger the poster. Every run of a
# listed workflow will trigger the poster, which will no-op if no
# `pr-review` artifact exists.
workflows:
- "Static Analysis"
types:
- completed
permissions:
pull-requests: write
actions: read
contents: read
jobs:
post:
name: Post PR Review
runs-on: ubuntu-latest
# Only run for pull_request producer runs. Push-to-main and other
# non-PR triggers have no review to post, so gating at the job level
# surfaces those as a clean "Skipped" in the UI instead of a
# silent no-op buried inside the script.
if: >-
github.event.workflow_run.conclusion != 'cancelled'
&& github.event.workflow_run.event == 'pull_request'
steps:
# Checkout runs first so the poster scripts are available AND so
# that actions/checkout@v6's default clean step does not delete the
# artifact zip that the next step writes into the workspace.
# Sparse-checkout restricts the materialized tree to just the
# poster script and its stdlib helper module.
- name: Checkout poster script only
uses: actions/checkout@v6
with:
sparse-checkout: |
Tools/ci/pr-review-poster.py
Tools/ci/_github_helpers.py
sparse-checkout-cone-mode: false
- name: Download pr-review artifact
id: download
uses: actions/github-script@v8
with:
script: |
const artifacts = await github.rest.actions.listWorkflowRunArtifacts({
owner: context.repo.owner,
repo: context.repo.repo,
run_id: context.payload.workflow_run.id,
});
const match = artifacts.data.artifacts.find(a => a.name === 'pr-review');
if (!match) {
core.info('No pr-review artifact on this run; nothing to post.');
core.setOutput('found', 'false');
return;
}
const download = await github.rest.actions.downloadArtifact({
owner: context.repo.owner,
repo: context.repo.repo,
artifact_id: match.id,
archive_format: 'zip',
});
const fs = require('fs');
fs.writeFileSync('pr-review.zip', Buffer.from(download.data));
core.setOutput('found', 'true');
- name: Unpack artifact
if: steps.download.outputs.found == 'true'
run: |
mkdir -p pr-review
unzip -q pr-review.zip -d pr-review
- name: Validate artifact
if: steps.download.outputs.found == 'true'
run: python3 Tools/ci/pr-review-poster.py validate pr-review
- name: Post PR review
if: steps.download.outputs.found == 'true'
env:
GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
run: python3 Tools/ci/pr-review-poster.py post pr-review

172
Tools/ci/_github_helpers.py Normal file
View File

@ -0,0 +1,172 @@
#!/usr/bin/env python3
"""
Shared GitHub REST helpers for PX4 CI scripts.
This module is imported by the PR poster scripts under Tools/ci/. It is
NOT an executable entry point; do not run it directly.
Provides:
- fail(msg) terminates the caller with a clear error
- GitHubClient(token) thin stdlib-only GitHub REST client with
single-request and paginated helpers
Python stdlib only. No third-party dependencies.
History: extracted from Tools/ci/pr-comment-poster.py so that
pr-comment-poster.py and pr-review-poster.py share the same HTTP plumbing
without duplicating ~100 lines of request/pagination/error-handling code.
"""
import json
import sys
import typing
import urllib.error
import urllib.request
GITHUB_API = 'https://api.github.com'
DEFAULT_USER_AGENT = 'px4-ci'
API_VERSION = '2022-11-28'
def fail(msg: str) -> typing.NoReturn:
"""Print an error to stderr and exit with status 1.
Annotated NoReturn so static checkers understand control does not
continue past a fail() call.
"""
print('error: {}'.format(msg), file=sys.stderr)
sys.exit(1)
def _parse_next_link(link_header):
"""Return the URL for rel="next" from an RFC 5988 Link header, or None.
The Link header is comma-separated entries of the form:
<https://...?page=2>; rel="next", <https://...?page=5>; rel="last"
We walk each entry and return the URL of the one whose rel attribute is
"next". Accept single-quoted rel values for robustness even though
GitHub always emits double quotes.
"""
if not link_header:
return None
for part in link_header.split(','):
segs = part.strip().split(';')
if len(segs) < 2:
continue
url_seg = segs[0].strip()
if not (url_seg.startswith('<') and url_seg.endswith('>')):
continue
url = url_seg[1:-1]
for attr in segs[1:]:
attr = attr.strip()
if attr == 'rel="next"' or attr == "rel='next'":
return url
return None
class GitHubClient:
"""Minimal GitHub REST client backed by the Python stdlib.
Each instance holds a token and a user-agent so callers do not have to
thread them through every call. Methods return parsed JSON (or None for
empty responses) and raise RuntimeError with the server response body on
HTTP errors, so CI logs show what the API actually objected to.
Usage:
client = GitHubClient(token, user_agent='px4-pr-comment-poster')
body, headers = client.request('GET', 'repos/{o}/{r}/pulls/123')
for item in client.paginated('repos/{o}/{r}/pulls/123/reviews'):
...
"""
def __init__(self, token, user_agent=DEFAULT_USER_AGENT):
if not token:
raise ValueError('GitHub token is required')
self._token = token
self._user_agent = user_agent
def request(self, method, path_or_url, json_body=None):
"""GET/POST/PATCH/PUT/DELETE a single API path or absolute URL.
`path_or_url` may be either a relative API path (e.g.
"repos/PX4/PX4-Autopilot/pulls/123") or an absolute URL such as the
next-page URL returned from paginated results. Relative paths are
prefixed with the GitHub API base.
Returns (parsed_json_or_none, headers_dict). Raises RuntimeError
on HTTP or transport errors.
"""
url = self._resolve(path_or_url)
return self._do_request(method, url, json_body)
def paginated(self, path, per_page=100):
"""GET a path and follow rel="next" Link headers.
Yields items from each page's JSON array. Bumps per_page to 100
(GitHub's max) so large result sets take fewer round-trips.
Raises RuntimeError if any page response is not a JSON array.
"""
url = self._resolve(path)
sep = '&' if '?' in url else '?'
url = '{}{}per_page={}'.format(url, sep, per_page)
while url is not None:
body, headers = self._do_request('GET', url, None)
if body is None:
return
if not isinstance(body, list):
raise RuntimeError(
'expected JSON array from {}, got {}'.format(
url, type(body).__name__))
for item in body:
yield item
url = _parse_next_link(headers.get('Link'))
def _resolve(self, path_or_url):
if path_or_url.startswith('http://') or path_or_url.startswith('https://'):
return path_or_url
return '{}/{}'.format(GITHUB_API.rstrip('/'), path_or_url.lstrip('/'))
def _do_request(self, method, url, json_body):
data = None
headers = {
'Authorization': 'Bearer {}'.format(self._token),
'Accept': 'application/vnd.github+json',
# Pin the API version so GitHub deprecations don't silently
# change the response shape under us.
'X-GitHub-Api-Version': API_VERSION,
'User-Agent': self._user_agent,
}
if json_body is not None:
data = json.dumps(json_body).encode('utf-8')
headers['Content-Type'] = 'application/json; charset=utf-8'
req = urllib.request.Request(
url, data=data, method=method, headers=headers)
try:
with urllib.request.urlopen(req) as resp:
raw = resp.read()
# HTTPMessage is case-insensitive on lookup but its items()
# preserves the original case. GitHub sends "Link" with a
# capital L, which is what _parse_next_link expects.
resp_headers = dict(resp.headers.items())
if not raw:
return None, resp_headers
return json.loads(raw.decode('utf-8')), resp_headers
except urllib.error.HTTPError as e:
# GitHub error bodies are JSON with a "message" field and often
# a "documentation_url". Dump the raw body into the exception so
# the CI log shows exactly what the API objected to. A bare
# "HTTP 422" tells us nothing useful.
try:
err_body = e.read().decode('utf-8', errors='replace')
except Exception:
err_body = '(no body)'
raise RuntimeError(
'GitHub API {} {} failed: HTTP {} {}\n{}'.format(
method, url, e.code, e.reason, err_body))
except urllib.error.URLError as e:
# Network layer failure (DNS, TLS, connection reset). No HTTP
# response to parse; just surface the transport reason.
raise RuntimeError(
'GitHub API {} {} failed: {}'.format(method, url, e.reason))

View File

@ -0,0 +1,539 @@
#!/usr/bin/env python3
#
# clang-tidy-fixes-to-review.py
#
# Producer-side helper that converts a clang-tidy fixes.yml file into a
# pr-review artifact (manifest.json + comments.json) suitable for
# Tools/ci/pr-review-poster.py.
#
# This script runs inside the clang-tidy job's px4-dev container so it can
# read the source tree directly and look up byte offsets in the original
# files. The output it writes is a fully-baked array of review comments;
# the poster never reads source files or fixes.yml.
#
# ----------------------------------------------------------------------------
# ATTRIBUTION
# ----------------------------------------------------------------------------
# This script reuses the diagnostic-to-review-comment translation logic
# from platisd/clang-tidy-pr-comments. The original work is:
#
# MIT License
#
# Copyright (c) 2021 Dimitris Platis
#
# Permission is hereby granted, free of charge, to any person obtaining a
# copy of this software and associated documentation files (the
# "Software"), to deal in the Software without restriction, including
# without limitation the rights to use, copy, modify, merge, publish,
# distribute, sublicense, and/or sell copies of the Software, and to
# permit persons to whom the Software is furnished to do so, subject to
# the following conditions:
#
# The above copyright notice and this permission notice shall be included
# in all copies or substantial portions of the Software.
#
# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
# IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
# CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
#
# Adapted parts:
# - get_diff_line_ranges_per_file() and its inner change_to_line_range()
# - generate_review_comments() and its nested helpers
# (get_line_by_offset, validate_warning_applicability,
# calculate_replacements_diff, markdown, markdown_url,
# diagnostic_name_visual, generate_single_comment)
# - reorder_diagnostics()
#
# Removed parts (handled by Tools/ci/pr-review-poster.py instead):
# - post_review_comments / dismiss_change_requests / resolve_conversations
# - the original argparse main and the requests-based HTTP layer
#
# Adaptation notes:
# - The HTTP layer is rewritten on top of Tools/ci/_github_helpers.py so
# this script does not depend on the third-party `requests` package.
# - Conversation resolution (the GraphQL path) is intentionally dropped
# for v1; revisit if it turns out to be missed.
# - Clang-Tidy 8 upconvert is preserved verbatim.
#
# ----------------------------------------------------------------------------
# Bounded assumptions (documented for future maintainers):
# - Source files are UTF-8 (we read them as latin_1, matching clang-tidy's
# own byte-offset model, and the offsets we surface are line counts)
# - Source files use LF line endings
# - Malformed entries in fixes.yml are skipped with a warning rather than
# crashing the job
#
# Dependencies: pyyaml + Tools/ci/_github_helpers.py.
# pyyaml is preinstalled in the px4-dev container; this script is intended
# to run there, not on bare ubuntu-latest.
"""Convert a clang-tidy fixes.yml into a pr-review artifact."""
import argparse
import difflib
import json
import os
import posixpath
import re
import sys
import urllib.parse
import yaml
import _github_helpers
from _github_helpers import fail as _fail
# Markers used inside the per-comment body to call out severity. Plain
# strings rather than emojis to keep the file emoji-free per project
# preferences; the rendered Markdown is unaffected.
SINGLE_COMMENT_MARKERS = {
'Error': '**[error]**',
'Warning': '**[warning]**',
'Remark': '**[remark]**',
'fallback': '**[note]**',
}
# ---------------------------------------------------------------------------
# Diff-range parsing (adapted from platisd)
# ---------------------------------------------------------------------------
def get_diff_line_ranges_per_file(pr_files):
"""Return a dict mapping each PR file path to a list of line ranges
(the +new-side hunks) parsed from its patch."""
def change_to_line_range(change):
split_change = change.split(',')
start = int(split_change[0])
size = int(split_change[1]) if len(split_change) > 1 else 1
return range(start, start + size)
result = {}
for pr_file in pr_files:
# Removed binary files etc. have no patch section.
if 'patch' not in pr_file:
continue
file_name = pr_file['filename']
# Match lines like '@@ -101,8 +102,11 @@'
git_line_tags = re.findall(
r'^@@ -.*? +.*? @@', pr_file['patch'], re.MULTILINE)
changes = [
tag.replace('@@', '').strip().split()[1].replace('+', '')
for tag in git_line_tags
]
result[file_name] = [
change_to_line_range(change) for change in changes
]
return result
def fetch_pull_request_files(client, repo, pr_number):
"""Yield file metadata objects for each file modified by the PR."""
path = 'repos/{}/pulls/{}/files'.format(repo, pr_number)
for entry in client.paginated(path):
yield entry
# ---------------------------------------------------------------------------
# Diagnostic ordering (adapted from platisd)
# ---------------------------------------------------------------------------
def reorder_diagnostics(diags):
"""Return diagnostics ordered Error -> Warning -> Remark -> other."""
errors = [d for d in diags if d.get('Level') == 'Error']
warnings = [d for d in diags if d.get('Level') == 'Warning']
remarks = [d for d in diags if d.get('Level') == 'Remark']
others = [
d for d in diags
if d.get('Level') not in {'Error', 'Warning', 'Remark'}
]
if others:
print(
'warning: some fixes have an unexpected Level (not Error, '
'Warning, or Remark)', file=sys.stderr)
return errors + warnings + remarks + others
# ---------------------------------------------------------------------------
# Comment generation (adapted from platisd)
# ---------------------------------------------------------------------------
def generate_review_comments(clang_tidy_fixes, repository_root,
diff_line_ranges_per_file,
single_comment_markers):
"""Yield review comment dicts for each clang-tidy diagnostic that
intersects the PR diff."""
def get_line_by_offset(file_path, offset):
# Clang-Tidy doesn't support multibyte encodings and measures
# offsets in bytes; latin_1 makes byte offsets and string offsets
# equivalent.
with open(repository_root + file_path, encoding='latin_1') as fh:
source = fh.read()
return source[:offset].count('\n') + 1
def validate_warning_applicability(file_path, start_line_num, end_line_num):
assert end_line_num >= start_line_num
for line_range in diff_line_ranges_per_file[file_path]:
assert line_range.step == 1
if (line_range.start <= start_line_num
and end_line_num < line_range.stop):
return True
return False
def calculate_replacements_diff(file_path, replacements):
# Apply replacements in reverse order so subsequent offsets do not
# shift.
replacements.sort(key=lambda item: (-item['Offset']))
with open(repository_root + file_path, encoding='latin_1') as fh:
source = fh.read()
changed = source
for replacement in replacements:
changed = (
changed[:replacement['Offset']]
+ replacement['ReplacementText']
+ changed[replacement['Offset'] + replacement['Length']:]
)
return difflib.Differ().compare(
source.splitlines(keepends=True),
changed.splitlines(keepends=True),
)
def markdown(s):
md_chars = '\\`*_{}[]<>()#+-.!|'
def escape_chars(s):
for ch in md_chars:
s = s.replace(ch, '\\' + ch)
return s
def unescape_chars(s):
for ch in md_chars:
s = s.replace('\\' + ch, ch)
return s
s = escape_chars(s)
s = re.sub(
"'([^']*)'",
lambda m: '`` ' + unescape_chars(m.group(1)) + ' ``',
s,
)
return s
def markdown_url(label, url):
return '[{}]({})'.format(label, url)
def diagnostic_name_visual(diagnostic_name):
visual = '**{}**'.format(markdown(diagnostic_name))
try:
first_dash_idx = diagnostic_name.index('-')
except ValueError:
return visual
namespace = urllib.parse.quote_plus(diagnostic_name[:first_dash_idx])
check_name = urllib.parse.quote_plus(
diagnostic_name[first_dash_idx + 1:])
return markdown_url(
visual,
'https://clang.llvm.org/extra/clang-tidy/checks/{}/{}.html'.format(
namespace, check_name),
)
def generate_single_comment(file_path, start_line_num, end_line_num,
name, message, single_comment_marker,
replacement_text=None):
result = {
'path': file_path,
'line': end_line_num,
'side': 'RIGHT',
'body': '{} {} {}\n{}'.format(
single_comment_marker,
diagnostic_name_visual(name),
single_comment_marker,
markdown(message),
),
}
if start_line_num != end_line_num:
result['start_line'] = start_line_num
result['start_side'] = 'RIGHT'
if replacement_text is not None:
if not replacement_text or replacement_text[-1] != '\n':
replacement_text += '\n'
result['body'] += '\n```suggestion\n{}```'.format(replacement_text)
return result
for diag in clang_tidy_fixes['Diagnostics']:
# Upconvert clang-tidy 8 format to 9+
if 'DiagnosticMessage' not in diag:
diag['DiagnosticMessage'] = {
'FileOffset': diag.get('FileOffset'),
'FilePath': diag.get('FilePath'),
'Message': diag.get('Message'),
'Replacements': diag.get('Replacements', []),
}
diag_message = diag['DiagnosticMessage']
diag_message['FilePath'] = posixpath.normpath(
(diag_message.get('FilePath') or '').replace(repository_root, ''))
for replacement in diag_message.get('Replacements') or []:
replacement['FilePath'] = posixpath.normpath(
replacement['FilePath'].replace(repository_root, ''))
diag_name = diag.get('DiagnosticName', '<unknown>')
diag_message_msg = diag_message.get('Message', '')
level = diag.get('Level', 'Warning')
single_comment_marker = single_comment_markers.get(
level, single_comment_markers['fallback'])
replacements = diag_message.get('Replacements') or []
if not replacements:
file_path = diag_message['FilePath']
offset = diag_message.get('FileOffset')
if offset is None:
print('warning: skipping {!r}: missing FileOffset'.format(
diag_name), file=sys.stderr)
continue
if file_path not in diff_line_ranges_per_file:
print(
"'{}' for {} does not apply to the files changed in "
'this PR'.format(diag_name, file_path))
continue
try:
line_num = get_line_by_offset(file_path, offset)
except (OSError, ValueError) as e:
print('warning: skipping {!r} on {}: {}'.format(
diag_name, file_path, e), file=sys.stderr)
continue
print("Processing '{}' at line {} of {}...".format(
diag_name, line_num, file_path))
if validate_warning_applicability(file_path, line_num, line_num):
yield generate_single_comment(
file_path,
line_num,
line_num,
diag_name,
diag_message_msg,
single_comment_marker=single_comment_marker,
)
else:
print('This warning does not apply to the lines changed '
'in this PR')
else:
for file_path in {item['FilePath'] for item in replacements}:
if file_path not in diff_line_ranges_per_file:
print(
"'{}' for {} does not apply to the files changed "
'in this PR'.format(diag_name, file_path))
continue
line_num = 1
start_line_num = None
end_line_num = None
replacement_text = None
try:
diff_iter = calculate_replacements_diff(
file_path,
[r for r in replacements if r['FilePath'] == file_path],
)
except (OSError, ValueError) as e:
print('warning: skipping {!r} on {}: {}'.format(
diag_name, file_path, e), file=sys.stderr)
continue
for line in diff_iter:
# Comment line, ignore.
if line.startswith('? '):
continue
# A '-' line is the start or continuation of a region
# to replace.
if line.startswith('- '):
if start_line_num is None:
start_line_num = line_num
end_line_num = line_num
else:
end_line_num = line_num
if replacement_text is None:
replacement_text = ''
line_num += 1
# A '+' line is part of the replacement text.
elif line.startswith('+ '):
if replacement_text is None:
replacement_text = line[2:]
else:
replacement_text += line[2:]
# A context line marks the end of a replacement region.
elif line.startswith(' '):
if replacement_text is not None:
if start_line_num is None:
# Pure addition: synthesize a one-line
# range and append the context line to
# the replacement.
start_line_num = line_num
end_line_num = line_num
replacement_text += line[2:]
print("Processing '{}' at lines {}-{} of {}...".format(
diag_name, start_line_num, end_line_num, file_path))
if validate_warning_applicability(
file_path, start_line_num, end_line_num):
yield generate_single_comment(
file_path,
start_line_num,
end_line_num,
diag_name,
diag_message_msg,
single_comment_marker=single_comment_marker,
replacement_text=replacement_text,
)
else:
print(
'This warning does not apply to the '
'lines changed in this PR')
start_line_num = None
end_line_num = None
replacement_text = None
line_num += 1
else:
# Unknown difflib prefix; skip rather than abort.
print('warning: unexpected diff prefix {!r}; '
'skipping diagnostic'.format(line[:2]),
file=sys.stderr)
break
# End of file with a pending replacement region.
if replacement_text is not None and start_line_num is not None:
print("Processing '{}' at lines {}-{} of {}...".format(
diag_name, start_line_num, end_line_num, file_path))
if validate_warning_applicability(
file_path, start_line_num, end_line_num):
yield generate_single_comment(
file_path,
start_line_num,
end_line_num,
diag_name,
diag_message_msg,
single_comment_marker=single_comment_marker,
replacement_text=replacement_text,
)
else:
print('This warning does not apply to the lines '
'changed in this PR')
# ---------------------------------------------------------------------------
# Entry point
# ---------------------------------------------------------------------------
def main(argv=None):
parser = argparse.ArgumentParser(
description='Convert a clang-tidy fixes.yml into a pr-review '
'artifact (manifest.json + comments.json).',
)
parser.add_argument('--fixes', required=True,
help='Path to fixes.yml from clang-tidy')
parser.add_argument('--repo-root', required=True,
help='Path to the repository root containing the '
'source files referenced by fixes.yml')
parser.add_argument('--repo', required=True,
help='owner/name of the repository')
parser.add_argument('--pr-number', required=True, type=int,
help='Pull request number')
parser.add_argument('--commit-sha', required=True,
help='40-char hex commit SHA the review will pin to')
parser.add_argument('--out-dir', required=True,
help='Directory to write manifest.json and '
'comments.json')
parser.add_argument(
'--marker',
default='<!-- pr-review-poster:clang-tidy -->',
help='Marker string embedded in the review body so the poster '
'can find and dismiss stale runs')
parser.add_argument(
'--event',
default='REQUEST_CHANGES',
choices=('COMMENT', 'REQUEST_CHANGES'),
help='GitHub review event type')
parser.add_argument(
'--summary', default='',
help='Optional review summary text appended to the review body')
args = parser.parse_args(argv)
if args.pr_number <= 0:
_fail('--pr-number must be > 0')
if not re.match(r'^[0-9a-f]{40}$', args.commit_sha):
_fail('--commit-sha must be a 40-char lowercase hex string')
token = os.environ.get('GITHUB_TOKEN')
if not token:
_fail('GITHUB_TOKEN is not set')
# Normalize the repo root with a trailing slash so the platisd-style
# str.replace() trick still strips it cleanly.
repo_root = args.repo_root
if not repo_root.endswith(os.sep):
repo_root = repo_root + os.sep
os.makedirs(args.out_dir, exist_ok=True)
client = _github_helpers.GitHubClient(token, user_agent='px4-clang-tidy-fixes-to-review')
print('Fetching PR file list from GitHub...')
pr_files = list(fetch_pull_request_files(client, args.repo, args.pr_number))
diff_line_ranges_per_file = get_diff_line_ranges_per_file(pr_files)
print('Loading clang-tidy fixes from {}...'.format(args.fixes))
if not os.path.isfile(args.fixes):
# No fixes file means clang-tidy ran cleanly. Emit an empty
# comments.json so the poster can short-circuit.
comments = []
else:
with open(args.fixes, encoding='utf-8') as fh:
clang_tidy_fixes = yaml.safe_load(fh)
if (not clang_tidy_fixes
or 'Diagnostics' not in clang_tidy_fixes
or not clang_tidy_fixes['Diagnostics']):
comments = []
else:
clang_tidy_fixes['Diagnostics'] = reorder_diagnostics(
clang_tidy_fixes['Diagnostics'])
comments = list(generate_review_comments(
clang_tidy_fixes,
repo_root,
diff_line_ranges_per_file,
single_comment_markers=SINGLE_COMMENT_MARKERS,
))
print('Generated {} review comment(s)'.format(len(comments)))
manifest = {
'pr_number': args.pr_number,
'marker': args.marker,
'event': args.event,
'commit_sha': args.commit_sha,
}
if args.summary:
manifest['summary'] = args.summary
manifest_path = os.path.join(args.out_dir, 'manifest.json')
comments_path = os.path.join(args.out_dir, 'comments.json')
with open(manifest_path, 'w', encoding='utf-8') as fh:
json.dump(manifest, fh, indent=2)
fh.write('\n')
with open(comments_path, 'w', encoding='utf-8') as fh:
json.dump(comments, fh, indent=2)
fh.write('\n')
print('Wrote {} and {}'.format(manifest_path, comments_path))
return 0
if __name__ == '__main__':
sys.exit(main())

View File

@ -37,9 +37,9 @@ import argparse
import json
import os
import sys
import typing
import urllib.error
import urllib.request
import _github_helpers
from _github_helpers import fail as _fail
# GitHub hard limit is 65535 bytes. Cap well under to leave headroom for
@ -53,7 +53,6 @@ MARKER_MAX_LEN = 200
ACCEPTED_MODES = ('upsert',)
GITHUB_API = 'https://api.github.com'
USER_AGENT = 'px4-pr-comment-poster'
@ -61,11 +60,6 @@ USER_AGENT = 'px4-pr-comment-poster'
# Validation
# ---------------------------------------------------------------------------
def _fail(msg: str) -> typing.NoReturn:
print('error: {}'.format(msg), file=sys.stderr)
sys.exit(1)
def _is_printable_ascii(s):
# Space (0x20) through tilde (0x7E) inclusive.
return all(0x20 <= ord(ch) <= 0x7E for ch in s)
@ -161,121 +155,11 @@ def validate_manifest(directory):
}
# ---------------------------------------------------------------------------
# GitHub HTTP helpers
# ---------------------------------------------------------------------------
def _github_request(method, url, token, json_body=None):
"""Perform a single GitHub REST request.
Returns a tuple (parsed_json_or_none, headers_dict). Raises RuntimeError
with the server response body on HTTP errors so CI logs show what
GitHub complained about.
"""
data = None
headers = {
'Authorization': 'Bearer {}'.format(token),
'Accept': 'application/vnd.github+json',
# Pin the API version so GitHub deprecations don't silently change
# the response shape under us.
'X-GitHub-Api-Version': '2022-11-28',
'User-Agent': USER_AGENT,
}
if json_body is not None:
data = json.dumps(json_body).encode('utf-8')
headers['Content-Type'] = 'application/json; charset=utf-8'
req = urllib.request.Request(url, data=data, method=method, headers=headers)
try:
with urllib.request.urlopen(req) as resp:
raw = resp.read()
# HTTPMessage is case-insensitive on lookup but its items() preserves
# the original case. GitHub sends "Link" with a capital L, which is
# what _parse_next_link expects.
resp_headers = dict(resp.headers.items())
if not raw:
return None, resp_headers
return json.loads(raw.decode('utf-8')), resp_headers
except urllib.error.HTTPError as e:
# GitHub error bodies are JSON with a "message" field and often a
# "documentation_url". Dump the raw body into the exception so the CI
# log shows exactly what the API objected to. A bare "HTTP 422"
# tells us nothing useful.
try:
err_body = e.read().decode('utf-8', errors='replace')
except Exception:
err_body = '(no body)'
raise RuntimeError(
'GitHub API {} {} failed: HTTP {} {}\n{}'.format(
method, url, e.code, e.reason, err_body))
except urllib.error.URLError as e:
# Network layer failure (DNS, TLS, connection reset). No HTTP response
# to parse; just surface the transport reason.
raise RuntimeError(
'GitHub API {} {} failed: {}'.format(method, url, e.reason))
def _parse_next_link(link_header):
"""Return the URL for rel="next" from an RFC 5988 Link header, or None.
The Link header is comma-separated entries of the form:
<https://...?page=2>; rel="next", <https://...?page=5>; rel="last"
We walk each entry and return the URL of the one whose rel attribute is
"next". Accept single-quoted rel values for robustness even though GitHub
always emits double quotes.
"""
if not link_header:
return None
for part in link_header.split(','):
segs = part.strip().split(';')
if len(segs) < 2:
continue
url_seg = segs[0].strip()
if not (url_seg.startswith('<') and url_seg.endswith('>')):
continue
url = url_seg[1:-1]
for attr in segs[1:]:
attr = attr.strip()
if attr == 'rel="next"' or attr == "rel='next'":
return url
return None
def github_api(method, path, token, json_body=None):
"""GET/POST/PATCH a single GitHub API path. Non-paginated."""
url = '{}/{}'.format(GITHUB_API.rstrip('/'), path.lstrip('/'))
body, _ = _github_request(method, url, token, json_body=json_body)
return body
def github_api_paginated(path, token):
"""GET a GitHub API path and follow rel="next" Link headers.
Yields items from each page's JSON array.
"""
url = '{}/{}'.format(GITHUB_API.rstrip('/'), path.lstrip('/'))
# GitHub defaults to per_page=30. Bump to 100 (the max) so a PR with a
# few hundred comments fetches in 3 or 4 round-trips instead of 10+.
sep = '&' if '?' in url else '?'
url = '{}{}per_page=100'.format(url, sep)
while url is not None:
body, headers = _github_request('GET', url, token)
if body is None:
return
if not isinstance(body, list):
raise RuntimeError(
'expected JSON array from {}, got {}'.format(
url, type(body).__name__))
for item in body:
yield item
url = _parse_next_link(headers.get('Link'))
# ---------------------------------------------------------------------------
# Comment upsert
# ---------------------------------------------------------------------------
def find_existing_comment_id(token, repo, pr_number, marker):
def find_existing_comment_id(client, repo, pr_number, marker):
"""Return the id of the first PR comment whose body contains marker, or None.
PR comments are issue comments in GitHub's data model, so we hit
@ -286,7 +170,7 @@ def find_existing_comment_id(token, repo, pr_number, marker):
appear in user-written prose.
"""
path = 'repos/{}/issues/{}/comments'.format(repo, pr_number)
for comment in github_api_paginated(path, token):
for comment in client.paginated(path):
body = comment.get('body') or ''
if marker in body:
return comment.get('id')
@ -308,24 +192,22 @@ def build_final_body(body, marker):
return '{}\n\n{}\n'.format(body.rstrip('\n'), marker)
def upsert_comment(token, repo, pr_number, marker, body):
def upsert_comment(client, repo, pr_number, marker, body):
final_body = build_final_body(body, marker)
existing_id = find_existing_comment_id(token, repo, pr_number, marker)
existing_id = find_existing_comment_id(client, repo, pr_number, marker)
if existing_id is not None:
print('Updating comment {} on PR #{}'.format(existing_id, pr_number))
github_api(
client.request(
'PATCH',
'repos/{}/issues/comments/{}'.format(repo, existing_id),
token,
json_body={'body': final_body},
)
else:
print('Creating new comment on PR #{}'.format(pr_number))
github_api(
client.request(
'POST',
'repos/{}/issues/{}/comments'.format(repo, pr_number),
token,
json_body={'body': final_body},
)
@ -364,8 +246,9 @@ def cmd_post(args):
_fail('GITHUB_REPOSITORY must be "owner/name", got {!r}'.format(repo))
try:
client = _github_helpers.GitHubClient(token, user_agent=USER_AGENT)
upsert_comment(
token=token,
client=client,
repo=repo,
pr_number=result['pr_number'],
marker=result['marker'],

View File

@ -0,0 +1,468 @@
#!/usr/bin/env python3
"""
PR review-comment poster for analysis workflows.
Sibling of Tools/ci/pr-comment-poster.py. Where pr-comment-poster.py posts
sticky issue-style PR comments, this script posts line-anchored review
comments on the "Files changed" tab. Use it for tools like clang-tidy that
want to flag specific lines instead of (or in addition to) a rollup
comment.
This script is invoked from the `PR Review Poster` workflow which runs on
`workflow_run` in the base repository context. It consumes a `pr-review`
artifact produced by an upstream analysis job and posts a fresh PR review
via the GitHub REST API, dismissing any stale review the same producer
left on a previous run.
Artifact contract (directory passed on the command line):
manifest.json
{
"pr_number": 12345, (required, int > 0)
"marker": "<!-- pr-review-poster:clang-tidy -->", (required, printable ASCII)
"event": "REQUEST_CHANGES", (required, "COMMENT" | "REQUEST_CHANGES")
"commit_sha": "0123456789abcdef0123456789abcdef01234567",(required, 40 hex chars)
"summary": "Optional review body text" (optional)
}
comments.json
JSON array of line-anchored review comment objects:
[
{"path": "src/foo.cpp", "line": 42, "side": "RIGHT",
"body": "..."},
{"path": "src/bar.hpp", "start_line": 10, "line": 15,
"side": "RIGHT", "start_side": "RIGHT", "body": "..."}
]
Note: an `APPROVE` event is intentionally NOT supported. Bots should never
approve a pull request.
Security: this script is run in a write-token context from a workflow that
MUST NOT check out PR code. Both manifest.json and comments.json are
treated as opaque data. The marker is validated to printable ASCII only
before use, and only reviews authored by github-actions[bot] whose body
contains the marker can be dismissed (a fork cannot spoof either).
Subcommands:
validate <dir> Validate that <dir> contains a conforming manifest +
comments file.
post <dir> Validate, then dismiss any stale matching review and
post a new review on the target PR. Requires env
GITHUB_TOKEN and GITHUB_REPOSITORY.
Python stdlib only. No third-party dependencies.
"""
import argparse
import json
import os
import re
import sys
import time
import _github_helpers
from _github_helpers import fail as _fail
USER_AGENT = 'px4-pr-review-poster'
# Marker length bounds. 1..200 is plenty for an HTML comment tag such as
# "<!-- pr-review-poster:clang-tidy -->".
MARKER_MIN_LEN = 1
MARKER_MAX_LEN = 200
# Cap per-comment body size well under GitHub's hard limit so we leave
# headroom for the wrapping JSON envelope. Empirically GitHub allows ~65535
# bytes per review comment body; 60000 is a safe ceiling.
MAX_COMMENT_BODY_BYTES = 60000
# Cap on number of comments per single review POST. platisd uses 10. The
# value matters because GitHub's review-creation endpoint has a payload
# size limit and review comments occasionally trip an abuse-detection
# threshold when posted in very large batches. Smaller chunks also let us
# spread the work across multiple reviews so a single bad entry only
# fails its own chunk.
COMMENTS_PER_REVIEW = 10
# Sleep between successive review POSTs to stay clear of GitHub's
# secondary rate limits. platisd uses 10s; 5s is enough for our volume
# and cuts user-visible latency.
SLEEP_BETWEEN_CHUNKS_SECONDS = 5
ACCEPTED_EVENTS = ('COMMENT', 'REQUEST_CHANGES')
ACCEPTED_SIDES = ('LEFT', 'RIGHT')
COMMIT_SHA_RE = re.compile(r'^[0-9a-f]{40}$')
# The login GitHub assigns to the built-in actions token. Used to filter
# the list of existing reviews so we never touch a human reviewer's review.
BOT_LOGIN = 'github-actions[bot]'
# ---------------------------------------------------------------------------
# Validation
# ---------------------------------------------------------------------------
def _is_printable_ascii(s):
return all(0x20 <= ord(ch) <= 0x7E for ch in s)
def validate_marker(marker):
"""Validate the marker string. See pr-comment-poster.py for rationale."""
if not isinstance(marker, str):
_fail('marker must be a string')
n = len(marker)
if n < MARKER_MIN_LEN or n > MARKER_MAX_LEN:
_fail('marker length out of range ({}..{}): {}'.format(
MARKER_MIN_LEN, MARKER_MAX_LEN, n))
if not _is_printable_ascii(marker):
_fail('marker contains non-printable or non-ASCII character')
def _validate_comment_entry(idx, entry):
"""Validate a single review-comment entry. Raises via _fail on error."""
if not isinstance(entry, dict):
_fail('comments[{}]: must be an object'.format(idx))
path = entry.get('path')
if not isinstance(path, str) or not path:
_fail('comments[{}].path: required non-empty string'.format(idx))
line = entry.get('line')
if not isinstance(line, int) or isinstance(line, bool) or line <= 0:
_fail('comments[{}].line: required positive integer'.format(idx))
side = entry.get('side', 'RIGHT')
if side not in ACCEPTED_SIDES:
_fail('comments[{}].side: must be one of {} (got {!r})'.format(
idx, ', '.join(ACCEPTED_SIDES), side))
if 'start_line' in entry:
start_line = entry['start_line']
if (not isinstance(start_line, int)
or isinstance(start_line, bool)
or start_line <= 0):
_fail('comments[{}].start_line: must be positive integer'.format(idx))
if start_line >= line:
_fail('comments[{}].start_line ({}) must be < line ({})'.format(
idx, start_line, line))
start_side = entry.get('start_side', side)
if start_side not in ACCEPTED_SIDES:
_fail('comments[{}].start_side: must be one of {}'.format(
idx, ', '.join(ACCEPTED_SIDES)))
body = entry.get('body')
if not isinstance(body, str) or not body:
_fail('comments[{}].body: required non-empty string'.format(idx))
body_bytes = len(body.encode('utf-8'))
if body_bytes > MAX_COMMENT_BODY_BYTES:
_fail('comments[{}].body too large: {} bytes (max {})'.format(
idx, body_bytes, MAX_COMMENT_BODY_BYTES))
def validate_manifest(directory):
"""Validate <directory>/manifest.json and <directory>/comments.json.
Returns a dict with keys: pr_number, marker, event, commit_sha,
summary, comments (list of validated comment dicts).
"""
manifest_path = os.path.join(directory, 'manifest.json')
comments_path = os.path.join(directory, 'comments.json')
if not os.path.isfile(manifest_path):
_fail('manifest.json missing at {}'.format(manifest_path))
if not os.path.isfile(comments_path):
_fail('comments.json missing at {}'.format(comments_path))
try:
with open(manifest_path, 'r', encoding='utf-8') as f:
manifest = json.load(f)
except (OSError, json.JSONDecodeError) as e:
_fail('manifest.json is not valid JSON: {}'.format(e))
if not isinstance(manifest, dict):
_fail('manifest.json must be a JSON object')
pr_number = manifest.get('pr_number')
if not isinstance(pr_number, int) or isinstance(pr_number, bool):
_fail('pr_number must be an integer')
if pr_number <= 0:
_fail('pr_number must be > 0 (got {})'.format(pr_number))
marker = manifest.get('marker')
validate_marker(marker)
event = manifest.get('event')
if event not in ACCEPTED_EVENTS:
_fail('event must be one of {} (got {!r}). APPROVE is intentionally '
'forbidden.'.format(', '.join(ACCEPTED_EVENTS), event))
commit_sha = manifest.get('commit_sha')
if not isinstance(commit_sha, str) or not COMMIT_SHA_RE.match(commit_sha):
_fail('commit_sha must be a 40-character lowercase hex string')
summary = manifest.get('summary', '')
if summary is None:
summary = ''
if not isinstance(summary, str):
_fail('summary must be a string if present')
try:
with open(comments_path, 'r', encoding='utf-8') as f:
comments = json.load(f)
except (OSError, json.JSONDecodeError) as e:
_fail('comments.json is not valid JSON: {}'.format(e))
if not isinstance(comments, list):
_fail('comments.json must be a JSON array')
for idx, entry in enumerate(comments):
_validate_comment_entry(idx, entry)
return {
'pr_number': pr_number,
'marker': marker,
'event': event,
'commit_sha': commit_sha,
'summary': summary,
'comments': comments,
}
# ---------------------------------------------------------------------------
# Stale-review dismissal
# ---------------------------------------------------------------------------
def find_stale_reviews(client, repo, pr_number, marker):
"""Yield (id, state) for each existing review owned by the bot AND
whose body contains the marker.
Filtering on BOTH author == github-actions[bot] AND marker-in-body is
the security invariant: a fork PR cannot impersonate the bot login,
and a fork PR also cannot inject the marker into a human reviewer's
body without API access.
"""
path = 'repos/{}/pulls/{}/reviews'.format(repo, pr_number)
for review in client.paginated(path):
user = review.get('user') or {}
if user.get('login') != BOT_LOGIN:
continue
body = review.get('body') or ''
if marker not in body:
continue
yield review.get('id'), review.get('state')
def dismiss_stale_reviews(client, repo, pr_number, marker):
"""Dismiss (or, for PENDING reviews, delete) every stale matching review."""
dismissal_message = 'Superseded by a newer run'
for review_id, state in find_stale_reviews(client, repo, pr_number, marker):
if review_id is None:
continue
if state == 'DISMISSED':
# Already inert; nothing to do.
continue
if state == 'PENDING':
# PENDING reviews cannot be dismissed; they must be deleted.
print('Deleting pending stale review {}'.format(review_id))
try:
client.request(
'DELETE',
'repos/{}/pulls/{}/reviews/{}'.format(
repo, pr_number, review_id))
except RuntimeError as e:
# Don't abort the run on dismissal failure: the new review
# will still be posted.
print('warning: failed to delete pending review {}: {}'.format(
review_id, e), file=sys.stderr)
continue
print('Dismissing stale review {} (state={})'.format(review_id, state))
try:
client.request(
'PUT',
'repos/{}/pulls/{}/reviews/{}/dismissals'.format(
repo, pr_number, review_id),
json_body={
'message': dismissal_message,
'event': 'DISMISS',
},
)
except RuntimeError as e:
print('warning: failed to dismiss review {}: {}'.format(
review_id, e), file=sys.stderr)
# ---------------------------------------------------------------------------
# Review posting
# ---------------------------------------------------------------------------
def _chunk(lst, n):
"""Yield successive n-sized slices of lst."""
for i in range(0, len(lst), n):
yield lst[i:i + n]
def _build_review_body(marker, summary, chunk_index, chunk_total):
"""Construct the review body text.
Always begins with the marker (so future runs can find and dismiss
this review). Appends a chunk index when the comment set is split
across multiple reviews, and the producer-supplied summary if any.
"""
parts = [marker]
if chunk_total > 1:
parts.append('({}/{})'.format(chunk_index + 1, chunk_total))
if summary:
parts.append('')
parts.append(summary)
return '\n'.join(parts)
def _comment_to_api(entry):
"""Project a validated comment dict to the GitHub API shape."""
api = {
'path': entry['path'],
'line': entry['line'],
'side': entry.get('side', 'RIGHT'),
'body': entry['body'],
}
if 'start_line' in entry:
api['start_line'] = entry['start_line']
api['start_side'] = entry.get('start_side', api['side'])
return api
def post_review(client, repo, pr_number, marker, event, commit_sha, summary,
comments):
"""Post one or more reviews containing the validated comments.
Comments are split into COMMENTS_PER_REVIEW-sized chunks. Each chunk
becomes its own review POST. A failed chunk logs a warning and the
loop continues to the next chunk.
"""
chunks = list(_chunk(comments, COMMENTS_PER_REVIEW))
total = len(chunks)
if total == 0:
print('No comments to post; skipping review creation.')
return
posted_any = False
for idx, chunk in enumerate(chunks):
if idx > 0:
time.sleep(SLEEP_BETWEEN_CHUNKS_SECONDS)
body = _build_review_body(marker, summary, idx, total)
payload = {
'commit_id': commit_sha,
'body': body,
'event': event,
'comments': [_comment_to_api(c) for c in chunk],
}
print('Posting review chunk {}/{} with {} comment(s)'.format(
idx + 1, total, len(chunk)))
try:
client.request(
'POST',
'repos/{}/pulls/{}/reviews'.format(repo, pr_number),
json_body=payload,
)
posted_any = True
except RuntimeError as e:
# Most common cause is HTTP 422: a comment refers to a line
# GitHub does not consider part of the diff. Skip the bad
# chunk and keep going so other findings still get posted.
print('warning: review chunk {}/{} failed: {}'.format(
idx + 1, total, e), file=sys.stderr)
if not posted_any:
# If every single chunk failed, surface that as a hard error so
# the workflow turns red and a human looks at it.
_fail('all review chunks failed to post; see warnings above')
# ---------------------------------------------------------------------------
# Entry points
# ---------------------------------------------------------------------------
def cmd_validate(args):
result = validate_manifest(args.directory)
print(('ok: pr_number={} marker_len={} event={} commit_sha={} '
'comments={} summary_len={}').format(
result['pr_number'],
len(result['marker']),
result['event'],
result['commit_sha'],
len(result['comments']),
len(result['summary']),
))
return 0
def cmd_post(args):
result = validate_manifest(args.directory)
# Empty comment lists short-circuit silently. A producer that ran but
# found nothing to flag should not generate noise on the PR.
if len(result['comments']) == 0:
print('No comments in artifact; nothing to post.')
return 0
token = os.environ.get('GITHUB_TOKEN')
if not token:
_fail('GITHUB_TOKEN is not set')
repo = os.environ.get('GITHUB_REPOSITORY')
if not repo:
_fail('GITHUB_REPOSITORY is not set (expected "owner/name")')
if '/' not in repo:
_fail('GITHUB_REPOSITORY must be "owner/name", got {!r}'.format(repo))
try:
client = _github_helpers.GitHubClient(token, user_agent=USER_AGENT)
dismiss_stale_reviews(
client=client,
repo=repo,
pr_number=result['pr_number'],
marker=result['marker'],
)
post_review(
client=client,
repo=repo,
pr_number=result['pr_number'],
marker=result['marker'],
event=result['event'],
commit_sha=result['commit_sha'],
summary=result['summary'],
comments=result['comments'],
)
except RuntimeError as e:
_fail(str(e))
return 0
def main(argv=None):
parser = argparse.ArgumentParser(
description='Validate and post line-anchored PR review comments '
'from CI artifacts.',
)
sub = parser.add_subparsers(dest='command', required=True)
p_validate = sub.add_parser(
'validate',
help='Validate manifest.json and comments.json in the given directory.',
)
p_validate.add_argument('directory')
p_validate.set_defaults(func=cmd_validate)
p_post = sub.add_parser(
'post',
help='Validate, then dismiss any stale review and post a new one. '
'Requires env GITHUB_TOKEN and GITHUB_REPOSITORY.',
)
p_post.add_argument('directory')
p_post.set_defaults(func=cmd_post)
args = parser.parse_args(argv)
return args.func(args)
if __name__ == '__main__':
sys.exit(main())