v0.2.0: gate-hardening — fail-open closes, codex parser currency, idempotent resolve - #2
Conversation
…y, idempotent resolve) Port four in-scope gate/parser/thread fixes from the internal pr-automation sibling and harden the thread classifier to be fail-closed by construction: - bot_authors: recognize the Codex GitHub reviewer (chatgpt-codex-connector / ^chatgpt-codex) so unresolved Codex threads are counted, not dropped — closes a MERGE_READY fail-open with an open Codex P1 thread. - resolve_bot_threads: replace the substring informational-classifier with a fail-closed pure-receipt check (NFKC-normalize + receipt-phrase span carve-out + Unicode-aware residual). Closes three adversarially-verified fail-open vectors (continuation-marker denylist bypass, non-Latin/homoglyph tokenizer blind spot, vocab-only negation). - parse_codex_output: parse codex CLI >=~0.13x native review ([P0-P3] markers + trailing em-dash path:start-end), alongside existing layouts. - graphql_helpers: add is_thread_resolved() + thread_reply_exists(); make reply_and_resolve() idempotent against double-fire (best-effort probes that never suppress the resolve). - bump version 0.1.0 -> 0.2.0; add CHANGELOG.md. Tests: 59 passed (was 37). ruff clean. leak gate clean (no home paths, gitleaks). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
![sourcery-ai[bot]](https://avatars.githubusercontent.com/in/48477?s=60&v=4){kind=small}
There was a problem hiding this comment.
Sorry @OrionArchitekton, you have reached your weekly rate limit of 500000 diff characters.
Please try again later or upgrade to continue using Sourcery
|
Warning Review limit reached
More reviews will be available in 5 minutes and 19 seconds. Learn how PR review limits work. Your organization has used up its prepaid credits, and credit purchases are no longer available. Enable the review add-on in the billing tab to keep reviews running — you're only billed for reviews past your plan's rate limits ($0.25/file). ⌛ How to resolve this issue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based credits. 🚦 How do rate limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please see our Fair Usage Limits Policy for further information. ℹ️ Review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (9)
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request hardens the merge admission control gate to prevent fail-open paths. Key changes include adding support for recognizing Codex reviewer bot authors, implementing idempotent thread reply and resolution in graphql_helpers.py, hardening the informational thread classifier in resolve_bot_threads.py to use a Unicode-aware pure-receipt check, and updating the parser to support newer Codex CLI native review formats. Feedback was provided on graphql_helpers.py to add or {} guards to prevent potential AttributeError exceptions if the GraphQL API returns a null data field.
Important
The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.
| reply_comment_id = ( | ||
| reply_result.get("data", {}) | ||
| .get("addPullRequestReviewThreadReply", {}) | ||
| .get("comment", {}) | ||
| .get("id") | ||
| ), | ||
| ) | ||
| reply_skipped = None | ||
|
|
||
| resolve_result = resolve_thread(thread_id) | ||
| return { | ||
| "thread_id": thread_id, | ||
| "reply_comment_id": reply_comment_id, | ||
| "resolved": ( | ||
| resolve_result.get("data", {}) | ||
| .get("resolveReviewThread", {}) | ||
| .get("thread", {}) | ||
| .get("isResolved", False) | ||
| ), |
There was a problem hiding this comment.
In is_thread_resolved and thread_reply_exists, the robust or {} pattern is used to guard against payload.get("data", {}) returning None (which is a standard GraphQL representation for partial or null data). However, in reply_and_resolve, this guard is missing when accessing reply_result and resolve_result. If the GraphQL API returns a null data field, this will raise an AttributeError: 'NoneType' object has no attribute 'get'. Adding the or {} fallback ensures consistency and prevents potential runtime crashes.
| reply_comment_id = ( | |
| reply_result.get("data", {}) | |
| .get("addPullRequestReviewThreadReply", {}) | |
| .get("comment", {}) | |
| .get("id") | |
| ), | |
| ) | |
| reply_skipped = None | |
| resolve_result = resolve_thread(thread_id) | |
| return { | |
| "thread_id": thread_id, | |
| "reply_comment_id": reply_comment_id, | |
| "resolved": ( | |
| resolve_result.get("data", {}) | |
| .get("resolveReviewThread", {}) | |
| .get("thread", {}) | |
| .get("isResolved", False) | |
| ), | |
| reply_comment_id = ( | |
| (reply_result.get("data", {}) or {}) | |
| .get("addPullRequestReviewThreadReply", {}) | |
| .get("comment", {}) | |
| .get("id") | |
| ) | |
| reply_skipped = None | |
| resolve_result = resolve_thread(thread_id) | |
| return { | |
| "thread_id": thread_id, | |
| "reply_comment_id": reply_comment_id, | |
| "resolved": ( | |
| (resolve_result.get("data", {}) or {}) | |
| .get("resolveReviewThread", {}) | |
| .get("thread", {}) | |
| .get("isResolved", False) | |
| ), | |
| "skipped": reply_skipped, | |
| } |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: d254a89cfd
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| _VERDICT_RE = re.compile(r"(?im)^verdict\s*:[ \t]*(.+?)[ \t]*$") | ||
| _FINDING_MARKER_RE = re.compile( | ||
| r"(?im)^[-*]\s*\[(critical|high|medium|low)\]\s*" | ||
| r"(?im)^[-*]\s*\[(critical|high|medium|low|P[0-3])\]\s*" |
There was a problem hiding this comment.
When the native output is the format documented just below as [P1] Title — path:start-end with no Verdict: line, this regex still requires a leading - or *. For that bare-marker output, parse_native_review_stdout extracts zero findings and returns schema_valid=False, so the new Codex format still becomes a parse error instead of producing fix requests.
Useful? React with 👍 / 👎.
1 participant
failclosed v0.2.0 — gate-hardening release
Ports four in-scope gate/parser/thread fixes from the internal
pr-automationsibling, and — after adversarial review — hardens the thread classifier beyond
its source. Two of the ports closed latent fail-open paths (a merge reported
MERGE_READYwhile a real, unresolved review signal was still open) — the exactfailure class failclosed exists to prevent.
Changes
bot_authors.yaml: recognize the Codex GitHub reviewer (chatgpt-codex-connector+^chatgpt-codex). Without it, unresolved Codex threads were dropped before the remaining-threads count → the gate could clear with an open Codex P1 thread.resolve_bot_threads.py: replace the substring informational-classifier with a fail-closed pure-receipt check (NFKC-normalize → carve out matched receipt phrases → any leftover alphanumeric content, in any script, keeps the thread open).parse_codex_output.py: parse codex CLI ≥~0.13x native review ([P1] Title — path:start-end):P0–P3markers + trailing em-dashpath:range.graphql_helpers.py:is_thread_resolved()+thread_reply_exists();reply_and_resolve()is now idempotent against double-fire (best-effort probes; never suppresses the resolve).Plus: version
0.1.0→0.2.0,CHANGELOG.md.Why the classifier went beyond a literal port
The ported classifier (pr-automation's "3-gate" version) was found by adversarial
review to still fail open three ways, fixed across three RED→GREEN rounds:
"No findings since v2 broke"auto-resolved;findings (
"No findings. Авторизация сломана",ＲＣＥ) and admitted its ownnegation (
"No findings. Action required.");fail-closed by a third adversarial pass. Every verified break is now a regression test.
Verification
pytest -q→ 59 passed (was 37 onmain).ruff check .→ clean.they surfaced (classifier fail-open) is fixed and re-confirmed CLOSED.
Deferred (pre-existing in the source; not introduced here)
Tracked, not silently dropped — all WARNING/INFO with low real-world trigger:
_FINDING_MARKER_REdoesn't match indented bullets orP4+; anapproveverdict co-occurring with a dropped finding could clear the gate(requires self-contradictory reviewer output).
reply_and_resolvecan double-post a comment if both idempotencyprobes error on a retry; exact-body dedup doesn't match across fix cycles (sha changes).
run_graphqlusesgh api graphql -F, which reads a@-prefixedvalue as a file — not reachable today (thread_id is a node ID, body is operator
text); switch to
-ffor defence-in-depth.classifies informational; non-exploitable (cannot encode an actionable finding).
No spec file: this is bugfix/port-class work (existing
test_*.pyseams are theexecutable contract), per the repo's spec-persistence rule.