test: skip Proxy-cost V8 microbench to stop noisy CI perf alerts - #9502
test: skip Proxy-cost V8 microbench to stop noisy CI perf alerts#9502lodekeeper wants to merge 1 commit into
Conversation
`packages/state-transition/test/perf/misc/proxy.test.ts` is a tight 100K-iteration hot loop measuring V8's Proxy.get overhead as a baseline reference. With no per-iteration work besides the property read, results are dominated by V8 inline-cache + engine-version effects and have very high run-to-run variance on the shared CI runner — `arrayWithProxy get 100000 times` tripped a 3.73x alert on PR ChainSafe#9494 with no Lodestar code change touching this path, and `proposeBlockBody type=full, size=empty` and `Full columns - reconstruct half of the blobs out of 20` have fired similarly on other unrelated PRs. No `src/` code uses `new Proxy(...)`, so this benchmark isn't tracking a real regression surface. Switch to `bench.skip(...)` in CI to stop generating false-positive alerts on unrelated PRs. The source stays in place so it can be flipped back to `bench(...)` locally for ad-hoc measurements. 🤖 Generated with AI assistance
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request disables the 'Proxy cost' microbenchmarks in CI by updating the benchmark calls to bench.skip. It also adds a detailed JSDoc comment explaining that these benchmarks are prone to high run-to-run variance on shared CI runners and do not track active production code paths. There are no review comments, so no additional feedback is provided.
Important
The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.
|
haven't seen benchmark failures due to these recently, closing |
2 participants
Summary
packages/state-transition/test/perf/misc/proxy.test.tsis a tight 100K-iteration hot loop measuring V8'sProxy.getoverhead as a baseline reference. With no per-iteration work besides the property read, results are dominated by V8 inline-cache and engine-version effects and have very high run-to-run variance on the shared CI runner.arrayWithProxy get 100000 timestripped a 3.73x alert on PR chore: stabilize fastConfirmation perf bench #9494 (issuecomment-4671463687) with no Lodestar code change touching this path.proposeBlockBody type=full, size=emptyandFull columns - reconstruct half of the blobs out of 20have fired similarly on chore: fix BlockInputSync metrics and logs #9481, fix: remove TOO_MANY_SKIPPED_SLOTS block error #9484, fix: boundSeenPayloadEnvelopeInputcache size to prevent OOM #9489 — same pattern of unrelated benchmarks tripping.No
src/code usesnew Proxy(...), so this benchmark isn't tracking a real regression surface. This PR switches the three Proxy-cost benches tobench.skip(...)so CI stops generating false-positive alerts on unrelated PRs. The source code stays in place so it can be flipped back tobench(...)locally for ad-hoc measurements (see the new docstring on thedescribeblock).Test plan
Benchmarkjob no longer emits alerts on the Proxy-cost suite.Type Checks/Unit Testsunaffected (skip doesn't touch the test discovery).🤖 Generated with Claude Code