Repository navigation
jc-batch-probe: refuse selections that name a resident plane (Codex follow-up to #1643) - #1645
Conversation
A Rel::Plane(i) selection was accepted and then failed inside the fold, because the probe passes no resident masks. Both paths now refuse it up front with BatchError::ResidentPlane; resident planes are U2's let-planes, and Cost does not count plane reads. Codex review on #1643. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VncKoyG6JkpsoqhqTGTv3f
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VncKoyG6JkpsoqhqTGTv3f
|
Important
This repository does not receive automatic reviews because it has fewer than 10 stars. ⚙️ Run configuration
Warning Billing warning: we have not been able to collect payment for this subscription for more than 72 hours. Please update the payment method or pay any pending invoices in Billing to avoid service interruption. 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 |
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 3321f57101
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".
…lanes The single in-order pass reported ResidentPlane for a batch whose plane came before a stale request, breaking the documented VersionMismatch precedence. Codex review on #1645. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VncKoyG6JkpsoqhqTGTv3f
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VncKoyG6JkpsoqhqTGTv3f
Follow-up to #1643. The fix for its Codex finding was pushed after the PR had already merged at
7c9ead98, so it did not reach main.Finding
A
Rel::Plane(i)selection was accepted byRequest::select, but the probe passes no resident masks, so it then failed inside the fold.Change
Both paths now refuse such a selection with
BatchError::ResidentPlanebefore anything runs, whether the plane stands alone or is nested insideAndorNot. Resident planes are U2's let-planes, andCostdoes not count plane reads, so supporting them here would understate the cost.Versions are checked over the whole batch first, so a batch with a stale request is refused as
VersionMismatchwherever a plane appears (Codex on this PR,7704c80e).Tests
-D warningsis clean.a_selection_naming_a_resident_plane_is_refusedcovers both forms on both paths.a_version_mismatch_wins_over_a_resident_plane_anywhere_in_the_batchplaces the plane in request 2 and the stale request in request 7.Not, and checking planes before versions in one pass.2026-10-20-u1-jc-batch.mdrecords the refusal, the precedence and the three disables.🤖 Generated with Claude Code
https://claude.ai/code/session_01VncKoyG6JkpsoqhqTGTv3f