GeezerVison

BugBot review

Generated by Claude Code

bugbot run

Generated by Claude Code

[burn-down agent] Labeled Ready for review at 826833aa34489ffed7c632724211a9ece3933c48.

• CI: 331/337 check runs green on this head (rest skipped by path/matrix filters), 0 failing. Mergeable; 7 commits behind main, only CHANGELOG.md overlaps and it merges cleanly.
• Bugbot: re-requested (the earlier run was cancelled) and it reviewed 826833aa34 with no issues found; no unresolved review threads.
• Reviewer focus: the chosen bounds (10 s connect, 60 s idle read, no total deadline) in api/retry.rs.

Generated by Claude Code

Reviewed final head 202e47744b96149f64a1f69a69c35e97420627a3.

The reproduced P2 JSON body-stall issue is fixed: all three JSON readers retain timeout/transport failures as Network with the underlying cause; completed malformed JSON stays Parse, and vendor retry behavior is preserved. 62 focused API tests passed.

The subsequent scan-performance failure was a stale Berry fixture, not a measured slowdown. Both main and this PR correctly rewrite package.json plus yarn.lock, while the fixture expected only yarn.lock; all 37 measured comparisons were unchanged. The branch now includes current main and reuses the isolated correction from #587, preserving provenance and requiring both files. The identical harness previously passed against the same baseline, and 29 benchmark tests pass locally. API source/tests are unchanged from the tested timeout fix.

Ready to merge as-is from this review. Final-head CI is complete: 332 successful checks, 6 skipped; no failures or pending checks. Bugbot passed, there are no unresolved review threads, and the PR is mergeable. No full local workspace or timing run was repeated.

BugBot review

Please review the JSON body timeout fix at 43269f46dcf0b6b663dfdd28bd38b48c1f1bd6bb. The reproduced issue and 62 passing focused tests are documented in the review comment above.

[agent] The scan performance check fails at 43269f4, and the cause is not in this PR. Job log.

• What fails: the yarn-berry/hosted and yarn-berry/rescan scenarios are marked INVALID for both binaries, including the base binary built from main @ 203e092. The error is redirect.rewrittenFiles: got ["package.json", "yarn.lock"], want ["yarn.lock"]. The other 38 scenarios are ≈, so this PR shows no timing change.
• Why it isn't this PR's: the bench suite (Add a scan benchmark suite and a CI performance gate #485, a79de97) landed before Fix yarn berry hosted pin leaking npm auth (#404) #465 (203e092), which changed yarn berry hosted mode so it also writes package.json. The fixture expectation at crates/socket-patch-bench/src/fixtures/npm.rs:626 still says &["yarn.lock"]. The head rescan's got [] probably comes from the base run's leftover fixture. This PR doesn't touch yarn or the bench.
• Fix: I couldn't find one in any open PR. Proposed patch, which belongs in its own PR rather than this one:

  -    Ok(fixture(&g, g.patches(true), &["yarn.lock"], &[]))
  +    Ok(fixture(&g, g.patches(true), &["package.json", "yarn.lock"], &[]))

  (in yarn_berry in crates/socket-patch-bench/src/fixtures/npm.rs). Please check the rescan scenario after that change. Every PR based on current main will fail this gate until it lands.
• Re-run: not done. The failure is deterministic on main's own binary, so a re-run would fail the same way.

Generated by Claude Code

BugBot review

Please review final commit 202e47744b96149f64a1f69a69c35e97420627a3. The JSON timeout fix is unchanged; this head also incorporates current main and the verified one-file Berry benchmark expectation correction. The PR description and review comment reflect the final state.