[FBumann]

Skips the metered macro runner on merges that cannot change what it measures.

Why

A release commit is a strict duplicate. release-please writes CHANGELOG.md and .release-please-manifest.json and touches no source — the git tag is the version — so CodSpeed spends a metered macro run reproducing the previous commit's number.

Measured against the last 40 merges to main: 17 would be skipped, 43%, and every one of them is a release commit.

SKIP  cf794c7  chore(main): release 0.0.1-alpha.89 (#572)
SKIP  839a50c  chore(main): release 0.0.1-alpha.88 (#569)
SKIP  15c9776  chore(main): release 0.0.1-alpha.87 (#567)
...
of the last 40 merges: 17 skipped, 23 measured
commits that would be skipped despite touching src/bench/pyproject/uv.lock: NONE

The rule, and what it deliberately excludes

Only what cannot change what the benchmark executes — narrower than "not code", and it keeps out three things that look obviously safe:

• pyproject.toml and uv.lock decide which polars gets installed, and a floor moving under us is exactly what this series is for
• bench/ is the workload
• codspeed.yml is the instrument

The failure mode to watch is not the release commit; it is a path added to this list later that turns out to reach the build, which then hides a regression in silence. The comment says so at the list.

paths-ignore skips only when every changed file matches, and merges here are squashes, so a pull request touching docs and src/ together is still measured.

The docs/** and **.md entries are principle rather than measurement: no docs-only merge appears in the last 40, so they save nothing yet and are included because they satisfy the same rule.

The cost, named

A skipped commit has no number, which the concurrency note calls a hole the next pull request falls into. That note is right in general and does not apply here: the hole sits at a commit whose executable tree is identical to its parent, so the fallback compares against the same code it would have.

What it does cost is legible — a pull request merged after one is footnoted unexpected-base. That footnote misled this session for a round trip on #520: it reads as "the baseline is stale, the comparison is unsafe" when the truth was "the baseline is the same code under a different sha". Hence the comment where somebody hitting it will look.

🤖 Generated with Claude Code

[codspeed]

Merging this PR will degrade performance by 3.09%

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 1 improved benchmark
❌ 1 regressed benchmark
✅ 12 untouched benchmarks
⏩ 38 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Memory	test_emit[profiled-s-lpspec-lp]	22.8 MB	26.2 MB	-12.9%
⚡	Memory	test_emit[sector-s-lpspec-highs]	4.5 MB	4.1 MB	+7.84%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing ci/codspeed-skips-the-release-commit (81db3f4) with main (cf794c7)}

Footnotes

1. 38 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩