You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
fix(review): a -diff lockfile makes both review lanes fail as partial #1109
Both hosted review lanes fail any pull request that adds or changes a package-lock.json. The fail class is partial: the publish step counts the lockfile as a binary file it cannot review, so it reports the head as not reviewed. The PR stays red on both review lanes whatever it contains.
##[error]claude-review: failure-class=partial: the review read only part of the pull request, so the head is not reviewed (read 452 of 453 changed file(s); 1 binary file(s) not reviewable as text: `plugins/formal-verification/skills/chart/scripts/package-lock.json`
The file is 1,459 bytes of JSON, not binary. claude-code-plugins .gitattributes lines 37-42 mark package-lock.json, packages.lock.json and *.lock as -diff ("Lockfiles: tracked, but suppress noisy diffs"). With -diff, git prints Binary files differ for them, and .github/actions/publish-review/publish.cjs counts that as an unreviewable binary.
Done when
A changed file the consumer marks -diff as a lockfile (or any -diff path that is text) does not make the review partial. Either the publish step exempts it from coverage and names it in the summary as not reviewed by design, or the review reads it with --text.
A test in .github/scripts/publish-review.test.cjs covers a -diff lockfile beside a real binary (an image), and only the image path's rule is unchanged.
claude-code-plugins PR #16458's review lanes pass on a rerun after the fix is released and pinned.
Closing as not planned: the hub misdiagnosed it. The partial review on claude-code-plugins PR 16458 came from a one-line sany.out no Read can show whole, not from the -diff lockfile, and binary_files never fails a review by itself (publish.cjs 237-243). Fixed in claude-code-plugins PR 16581 through generated-data. PR 1110 is closed.
Activity
kyle-sexton commentedon Nov 16, 2026
Closing as not planned: the hub misdiagnosed it. The partial review on claude-code-plugins PR 16458 came from a one-line sany.out no Read can show whole, not from the -diff lockfile, and binary_files never fails a review by itself (publish.cjs 237-243). Fixed in claude-code-plugins PR 16581 through generated-data. PR 1110 is closed.