Skip to content

Expire stale carried finding evidence#104

Closed
morgaesis wants to merge 4 commits into
mainfrom
codex/fix-stale-carried-findings
Closed

Expire stale carried finding evidence#104
morgaesis wants to merge 4 commits into
mainfrom
codex/fix-stale-carried-findings

Conversation

@morgaesis

Copy link
Copy Markdown
Contributor

Separate failed, bounded, and exhaustive review trust so completed bounded reviews expire changed or removed citations without dropping unchanged evidence outside the selected batches. Exact evidence remapping now covers duplicate lines and renames, with full and incremental regressions; the package version is 0.7.7 for the corrective release.

@postil-dev postil-dev Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

info 2 inline findings

Review details

Comment thread src/review.rs
Comment on lines +1519 to +1525
review_trust = if batch_failed {
filter::ReviewTrust::Failed
} else if risk_selected_review {
filter::ReviewTrust::Bounded
} else {
filter::ReviewTrust::Exhaustive
};

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

warn Bounded reviews now expire prior findings instead of carrying them as failing
warn · confidence 0.82 · kind: humanEscalation

The old code set full_review_trustworthy=false for bounded reviews, which carried prior findings forward and kept the gate failing. The new code assigns ReviewTrust::Bounded and the updated e2e test confirms prior findings from non-selected batches are now moved to resolved with gate.failing=false (tests/e2e.rs lines 2457, 2466-2468, 2480, 2488-2489). This means a real issue in a file outside the bounded-review selection is silently resolved and no longer blocks the merge. Confirm with the review-policy owner that auto-resolving out-of-scope prior findings is intended, and verify the reconcile implementation only expires findings whose cited lines are truly absent from the complete current diff, not merely absent from the selected batch.

Comment thread tests/e2e.rs
.args(["--output", "json"])
.assert()
.code(1);
.code(0);

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

warn Bounded reviews now drop prior findings from non-selected batches
warn · confidence 0.72 · kind: humanEscalation

The test was renamed from carrying changed prior findings to expiring them, and the expected exit code changed from 1 to 0, confirming that prior findings from batches outside the bounded selection are no longer carried forward. Users who rely on bounded reviews will silently lose baseline findings for files not in the current selection. Confirm with the product owner that expiring rather than carrying these findings is the intended behavior, and verify that the reconcile path in src/filter.rs line 292 (trust == ReviewTrust::Exhaustive) correctly gates this for both bounded and full review scopes.

@morgaesis

Copy link
Copy Markdown
Contributor Author

Superseded by #105, which contains the same reviewed fix plus the renamed-evidence regression on a clean PR baseline.

@morgaesis morgaesis closed this Jul 20, 2026
@morgaesis

Copy link
Copy Markdown
Contributor Author

@postil approve 6d3927b16f3b8f60c7fb78a24bd40e836d2cb94563019a966bee44096391ea70 -- Reviewed the bounded-review evidence lifecycle and accept this scoped compatibility tradeoff.

1 similar comment
@morgaesis

Copy link
Copy Markdown
Contributor Author

@postil approve 6d3927b16f3b8f60c7fb78a24bd40e836d2cb94563019a966bee44096391ea70 -- Reviewed the bounded-review evidence lifecycle and accept this scoped compatibility tradeoff.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant