Repository navigation
lab(drf-w3): check an unordered write against its whole facet group - #1768
Conversation
…ue of its facet Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G22yT6htkcdyXsihxxXdrg
Codex P2 on #1766 (merged): stage 1, stage_at 1@5, stage_at 2@6 passed, because after the sort the unordered write only neighbours the equal position-5 write. flush now carries the group's unordered value and refuses any positioned value that differs from it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G22yT6htkcdyXsihxxXdrg
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01G22yT6htkcdyXsihxxXdrg
|
Important
This repository does not receive automatic reviews because it has fewer than 10 stars.
|
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. |
This fixes the Codex P2 finding on #1766, which merged before the review finished.
Bug. Staging
stage(slot, facet, 1),stage_at(slot, facet, 1, 5)andstage_at(slot, facet, 2, 6)should refuse. Nothing orders the position-less write against the value-2 write, soflushhas no basis for choosing between them. Instead it passed and wrote 2. After the sort, the position-less write sits next to the equal-valued position-5 write only, and the check compared neighbours only.Fix. Within one row and field,
flushnow remembers the value of the position-less writes. It refuses any positioned write whose value differs from it. Writes with no position sort first, so all of them have been seen before the first positioned write.Tests. 18 pass on native and on
x86-64-v3; clippy-D warningsis clean.w3_an_unordered_write_conflicts_with_every_different_positioned_value_in_its_groupcovers Codex's case and refuses it.The board entry
2026-10-25-drf-w3-ordered-apply.mdrecords the case.Generated by Claude Code