Repository navigation
fix(baseline): Split release_post.py's gh api output on newlines only - #298
Conversation
Closes #297 Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
There was a problem hiding this comment.
Claude Review
PR #298 changes GitHub._list in the release_post.py template to split gh api output on "\n" instead of splitlines(), and adds a regression test. I found nothing to hold the merge.
jq's @JSON leaves U+2028 and U+2029 unescaped, so a PR file patch containing either character was cut into two non-JSON halves by splitlines(). Splitting on "\n" fixes that.
- Line endings: subprocess.run(text=True) already normalises \r and \r\n to \n, and @JSON escapes any \r and \n inside a record, so one record per line still holds.
- Other callers: commits(), files() and releases() all go through _list, so every record type is covered. Blank lines are still skipped by the line.strip() check.
- Test: it writes the separators raw with ensure_ascii=False, and the stub lambda path, jq=None matches how _list calls _api(path, jq). Under the old code splitlines() would split the first record at U+2028 and U+2029, and json.loads would raise. I did not check that json and REPO are defined in the test module; they are used by other tests in that file.
Not reviewed: none. I read the one diff piece and the release_post.py context pieces 001 to 003. I did not read context pieces 004 to 007 or the test file's own context.
Advisory
- commit_subject (release_post.py:103) and first_paragraph (around line 1458) still use splitlines() on untrusted text. A U+2028 or U+0085 in a commit subject would truncate the subject at that character. This only shortens text and loses or mis-parses nothing. You may want it to match the new "newlines only" rule.
- This is the template copy under "Support For New Projects/". If the repo also keeps a deployed copy (for example .github/scripts/release_post.py), it needs the same change.
No findings.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
There was a problem hiding this comment.
Claude Review
PR #298 makes GitHub._list split on "\n" only and makes commit_subject cut only at CRLF, CR or LF, so a U+2028 or U+2029 no longer splits a JSON record or a commit subject. The new tests are correct and would fail on the old code. The two edits themselves are sound. The same splitting bug still exists in update_blog_index, which the PR leaves unchanged; I report it as one finding. Advisory: first_paragraph also uses .splitlines() on post text, but it re-joins the lines with spaces, so the effect is cosmetic. All diff and context pieces were read, so nothing is left unreviewed. I did not check whether other copies of release_post.py in the repo need the same edit.
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
There was a problem hiding this comment.
Claude Review
The PR replaces str.splitlines() with splits on CRLF/CR/LF only in three places in release_post.py (GitHub._list, commit_subject, update_blog_index), so U+2028 and U+2029 no longer cut records, commit subjects or blog index rows. A lone CR now ends a commit subject, as Markdown treats it as a line break. Tests were added for each change. I traced the empty, whitespace-only, leading-newline, CRLF and trailing-CR cases and found no bug, so the verdict is no blockers. Not reviewed: I read only the first 767 lines of release_post.py (the rest, including the body of update_blog_index beyond the changed hunk, was not read) and not the test file's imports and fixtures (json, pytest, REPO, make_repo), which I assume exist. Both diff pieces were read in full. Advisory: other splitlines() calls may remain elsewhere in release_post.py (sanitizer, README/docs table handling) with the same U+2028/U+2029 exposure; I could not check.
No findings.
## Why Re-Apply the repo-ci-baseline Template so this repo carries the Template changes from mpaulosky/dotfiles#298 and mpaulosky/dotfiles#300. ## What changed - `c3b2cab` Apply commit, `apply.sh` output alone: brings mpaulosky/dotfiles#298 and mpaulosky/dotfiles#300. - `.github/scripts/post_claude_review.py` - `.github/scripts/release_post.py` - `.github/scripts/tests/test_post_claude_review.py` - `.github/scripts/tests/test_release_post.py` - `.github/workflows/claude-review.yml` - `.sandcastle/lib/branches.mts` - `.sandcastle/lib/branches.test.mts` - `.sandcastle/lib/build.mts` - `.sandcastle/lib/build.test.mts` ## Verification - The Apply commit is exactly `apply.sh`'s output on a branch up to date with `main`. - The pre-push gate (`scripts/gate.sh`) passed on `c3b2cab`. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
## Why Re-Apply the repo-ci-baseline Template so this repo carries the Template changes from mpaulosky/dotfiles#298 and mpaulosky/dotfiles#300. ## What changed - `726ce26` Apply commit, `apply.sh` output alone: brings mpaulosky/dotfiles#298 and mpaulosky/dotfiles#300. - `.github/scripts/post_claude_review.py` - `.github/scripts/release_post.py` - `.github/scripts/tests/test_post_claude_review.py` - `.github/scripts/tests/test_release_post.py` - `.github/workflows/claude-review.yml` - `.sandcastle/lib/branches.mts` - `.sandcastle/lib/branches.test.mts` - `.sandcastle/lib/build.mts` - `.sandcastle/lib/build.test.mts` ## Verification - The Apply commit is exactly `apply.sh`'s output on a branch up to date with `main`. - The pre-push gate (`scripts/gate.sh`) passed on `726ce26`. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
## Why Re-Apply the repo-ci-baseline Template so this repo carries the Template changes from mpaulosky/dotfiles#298 and mpaulosky/dotfiles#300. ## What changed - `dd366da` Apply commit, `apply.sh` output alone: brings mpaulosky/dotfiles#298 and mpaulosky/dotfiles#300. - `.github/scripts/post_claude_review.py` - `.github/scripts/release_post.py` - `.github/scripts/tests/test_post_claude_review.py` - `.github/scripts/tests/test_release_post.py` - `.github/workflows/claude-review.yml` ## Verification - The Apply commit is exactly `apply.sh`'s output on a branch up to date with `main`. - The pre-push gate (`scripts/gate.sh`) passed on `dd366da`. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
## Why Re-Apply the repo-ci-baseline Template so this repo carries the Template changes from mpaulosky/dotfiles#298 and mpaulosky/dotfiles#300. ## What changed - `aa7ef05` Apply commit, `apply.sh` output alone: brings mpaulosky/dotfiles#298 and mpaulosky/dotfiles#300. - `.github/scripts/post_claude_review.py` - `.github/scripts/release_post.py` - `.github/scripts/tests/test_post_claude_review.py` - `.github/scripts/tests/test_release_post.py` - `.github/workflows/claude-review.yml` - `.sandcastle/lib/branches.mts` - `.sandcastle/lib/branches.test.mts` - `.sandcastle/lib/build.mts` - `.sandcastle/lib/build.test.mts` ## Verification - The Apply commit is exactly `apply.sh`'s output on a branch up to date with `main`. - The pre-push gate (`scripts/gate.sh`) passed on `aa7ef05`. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: Teqslamer <teqslamer@gmail.com> Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
release_post.py'sGitHub._listreadgh api --jq '.[] | @json'output withstr.splitlines(). jq leaves U+2028 and U+2029 unescaped andsplitlines()splits on both, so a record whose text held one was cut in two andjson.loadsraisedJSONDecodeError. That crashed the Release workflow'sdocsjob on mpaulosky/IssueManager#328, so it got no release-blog PR._listnow splits on"\n"only. The new test puts both separators in a record and fails against the old code.update_blog_indexreadsdocs/blogs/README.mdback by CRLF, CR and LF only, so a row whose title holds U+2028 or U+2029 is kept.commit_subjectnow cuts a commit message at Markdown's line endings (CRLF, CR, LF) only, so a U+2028 or U+2029 stays in the subject instead of truncating it. A lone CR still ends the subject, because Markdown reads it as a line ending. Tests cover that and the rendered list item.Closes #297
🤖 Generated with Claude Code