[felickz]

Stacked on #240 (which is stacked on #238). Base is felickz-hotspots-oom-codeql-2-27-3, so the diff here is only the ci.yml change.

Problem

On the #240 CI run, all 13 compile-and-test jobs passed and every reported test passed, but "Validate test results" failed with jq: parse error: Invalid numeric literal at line 1, column 8.

The bad file was javascript-test-results/test_report_slice_1_of_4.json. Its first line was a JVM warning, not JSON:

[0.001s][warning][perf,memops] Cannot use file /tmp/hsperfdata_runner/2405 because it is locked by another process (errno = 11)

ci.yml starts four codeql test run --format=json processes at once and redirects each one's stdout directly into its report file. The JVM writes its own warnings to stdout too, so anything it prints ahead of the JSON corrupts the report. I scanned the test reports from 6 recent CI runs (about 150 files) and this was the only polluted one, so it is rare, but nothing in the workflow tolerates it.

Changes

1. Add -J=-XX:-UsePerfData to the codeql test run command. This stops the JVM from creating the per-PID /tmp/hsperfdata_<user>/<pid> file whose lock collided. The CLI accepts -J=<opt> (confirmed locally).
2. In "Validate test results", strip everything before the opening [ of each report (a lone [, or [] for an empty report) before running jq. A ::warning:: is emitted when something was dropped, and the step still fails if the remainder is not valid JSON or contains a failing test.

Testing

I ran the exact validate step from the workflow file under bash against sample reports:

Case	Expected	Result
clean report	pass	pass
warning line before JSON, plus a second clean slice	pass with warning	pass with warning
empty [] report	pass	pass
warning line before a report with a failing test	fail	fail, failing test printed
non-JSON garbage	fail	fail, "is not valid JSON"
no reports	pass	pass

The -J flag was only checked for CLI acceptance. Its effect on codeql test run is exercised by this PR's own CI run.

Notes

• This is independent of the hotspots fix in fix(hotspots): avoid cartesian product in getImportStringFrom (stacked on #238) #240. It could also be retargeted to main on its own.
• ci.yml only triggers for PRs targeting main. See the check status on this PR to confirm whether it runs when stacked.

[Copilot]

🔵 Needs a closer look

The stacked PR has no reported end-to-end CI checks exercising the modified CodeQL invocation and report validation.

No open issues 🎉

What changed in this PR

This PR hardens CodeQL CI report handling against JVM warnings that can corrupt JSON output and fail validation despite passing tests.

Changes:

• Disables JVM performance-data files for codeql test run to prevent the observed lock warning.
• Removes leading output before the report’s array opener, emits a warning when output is removed, and retains JSON and test-failure validation.

File	Description
.github/​workflows/​ci.yml	Updates the CodeQL JVM option and sanitizes reports before validation.

_{Review effort: Balanced}