#24436 Web checks: vp fmt's closing file count is not a finding the change adds

Fixes a false red in the webviews checks (fast gates and ci-web): vp fmt --check closes with Finished in 528ms on 2481 files using 4 threads.. When the base already has format findings and the pull request adds or removes any file, the merge and the base close on different file counts, and that closing line reads as the one finding the merge added.

Seen on

#24274's fast gates, twice, against two bases:

• run 39874062403 (base bb02624f): 1 finding(s) this change adds over the base: Finished in 2057ms on 2461 files using 4 threads.
• run 39891007201 (base da65a72f): 1 finding(s) this change adds over the base: Finished in 528ms on 2481 files using 4 threads.

Both runs say the other 3 vp fmt findings are on the base too, so the pull request added no format finding.

Fix

NOISE in scripts/ci/fast_gate_webviews_checks.py also matches Finished in … on N files, the same way it already skips the type check's Found N errors and N warnings summary. A file vp fmt names is still a finding.

Test

test_replay_of_24274_the_format_check_file_count_is_not_a_finding (huge-blue):

• At base code: fails with ['Finished in 528ms on 2481 files using 4 threads.'] != [].
• With the fix: passes. It also checks that a newly unformatted file is still charged.
• WebviewsChecksGate: all 22 tests pass.

risk:low. CI gate only; no app code.

Agent session: cosmic-gray

🤖 Generated with Claude Code

---

^{Need help on this PR? Tag @codesmith-bot with what you need.}

---

Summary by cubic

Fixes a false red in the webviews fast gates and CI-web: vp fmt --check's closing line (e.g. Finished in 528ms on 2481 files using 4 threads.) is no longer counted as a finding. When the base already has format findings and the pull request adds or removes any file, the merge and base close on different file counts, which previously read as the one finding the merge added (#24274).

• Treats only the whole closing file-count line as run noise, like the type check's Found N errors and N warnings summary; files the formatter actually names are still findings, including files whose names start like the summary.
• Adds a replay test covering #24274's case, confirming the summary is not charged and that a newly unformatted file, or a file named like the summary, is still reported.