feat: add MacOS support

[not-matthias]

No description provided.

[codspeed-hq]

Merging this PR will degrade performance by 16.23%

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

⚠️ 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

⚡ 7 improved benchmarks
❌ 12 regressed benchmarks
✅ 206 untouched benchmarks
🆕 1 new benchmark

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	WallTime	iterative fibo 15	312 ns	900 ns	-65.33%
❌	WallTime	iterative fibo 20	360 ns	900 ns	-60%
❌	WallTime	short body	1.7 µs	2.8 µs	-37.66%
❌	WallTime	short body	1.7 µs	2.6 µs	-33.64%
❌	WallTime	short body	1.8 µs	2.5 µs	-31.13%
❌	WallTime	short body	1.8 µs	2.5 µs	-30%
❌	WallTime	short body	1.7 µs	2.4 µs	-28.86%
❌	WallTime	short body	1.8 µs	2.5 µs	-27.96%
❌	Simulation	switch 2	8.3 µs	11.2 µs	-25.47%
❌	Memory	short body	408 B	536 B	-23.88%
❌	WallTime	fibo 10	2.1 µs	2.8 µs	-23.04%
❌	Simulation	wait 1ms	20 µs	22.8 µs	-11.95%
⚡	Memory	wait 1ms	16 B	7 B	×2.3
⚡	Simulation	short body	121.2 µs	76.1 µs	+59.37%
⚡	Simulation	long body	382.1 µs	336.1 µs	+13.69%
⚡	WallTime	switch 1	324 ns	288 ns	+12.5%
⚡	WallTime	test sync baz 10	360 ns	324 ns	+11.11%
⚡	WallTime	test sync baz 100	480 ns	432 ns	+11.11%
⚡	WallTime	test sync baz 10	384 ns	348 ns	+10.34%
🆕	WallTime	fibo darwin	N/A	7.9 ms	N/A

Tip

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

---

_{Comparing cod-2720-bump-instrument-hooks-in-codspeed-node-to-support-macos (c184ad6) with main (4dae798)}

[greptile-apps]

Greptile Summary

This PR adds macOS support to the CodSpeed Node.js benchmarking plugins by centralizing the __codspeed_root_frame__ wrapper into a shared wrapWithRootFrame/wrapWithRootFrameSync utility in @codspeed/core, and introduces benchmark timeline markers (BenchmarkStart/BenchmarkEnd) so the macOS walltime runner can bracket perf samples. CI is updated to remove the old walltime-macos-test job and replace it with a proper codspeed-walltime-macos job in codspeed.yml.

• packages/core/src/rootFrame.ts (new): Centralizes the __codspeed_root_frame__ wrapper that was previously inlined in every plugin; all four plugins now import and use it.
• packages/tinybench-plugin/src/shared.ts and vitest-plugin/src/walltime/index.ts: Add marker emission (addMarker with BenchmarkStart/BenchmarkEnd timestamps) so the runner can identify the benchmark window inside a profiling sample — critical for macOS walltime support.

Confidence Score: 4/5

The macOS walltime support is functionally sound in the vitest plugin path, but the tinybench plugin emits benchmark markers after closing the sample window, which both breaks the new integration test and produces incorrect profiling data on the tinybench walltime path.

In packages/tinybench-plugin/src/shared.ts, sendBenchmarkMarkers() is called after stopBenchmark() and setExecutedBenchmark(). The code's own comment, the vitest-plugin reference implementation, and the newly added integration test all require the markers to be emitted before stopBenchmark(). The test will fail as written, and any tinybench walltime benchmark run on macOS will have misattributed or missing flame graph data.

packages/tinybench-plugin/src/shared.ts — both wrapWithInstrumentHooks and wrapWithInstrumentHooksAsync call sendBenchmarkMarkers after stopBenchmark(), inverting the required ordering.

Important Files Changed

Filename	Overview
packages/tinybench-plugin/src/shared.ts	Adds marker emission to wrapWithInstrumentHooks/wrapWithInstrumentHooksAsync, but sendBenchmarkMarkers is called after stopBenchmark(), inverting the required ordering and breaking the newly added integration test.
packages/tinybench-plugin/tests/index.integ.test.ts	Adds test that asserts markers land between startBenchmark and stopBenchmark — but the current shared.ts implementation violates this invariant, so the test will fail.
packages/vitest-plugin/src/walltime/index.ts	Refactors tinybench hook to use wrapWithRootFrame; markers are correctly emitted before stopBenchmark(), restores task.fn in finally, and captures pid once outside the hot path.
packages/core/src/rootFrame.ts	New utility cleanly centralizes the codspeed_root_frame wrapper; both sync and async variants are well-typed and correct.
.github/workflows/codspeed.yml	Adds codspeed-walltime-macos job; previous threads already flagged the staging URL, diagnostic step, and mutable branch pin.
packages/tinybench-plugin/src/walltime.ts	Simplifies wrapTaskFunction to use shared wrapWithRootFrame/wrapWithRootFrameSync; logic is unchanged.
packages/vitest-plugin/src/analysis.ts	Swaps inline codspeed_root_frame for the shared helper; no behavioral change.
packages/benchmark.js-plugin/src/index.ts	Uses shared wrapWithRootFrame/wrapWithRootFrameSync; test mock updated to pass through the real implementations.
packages/vitest-plugin/benches/macos.bench.ts	New macOS-only benchmark, skipped via describe.skipIf on non-Darwin platforms.

Sequence Diagram

sequenceDiagram
    participant Plugin as Plugin (tinybench/vitest/benchmark.js)
    participant Core as @codspeed/core
    participant Runner as CodSpeed Runner

    Note over Plugin,Runner: Correct order (vitest walltime / this PR intent)
    Plugin->>Core: startBenchmark()
    Plugin->>Core: currentTimestamp() → runStart
    Plugin->>Plugin: execute benchmark fn() under __codspeed_root_frame__
    Plugin->>Core: currentTimestamp() → runEnd
    Plugin->>Core: addMarker(BenchmarkStart, runStart)
    Plugin->>Core: addMarker(BenchmarkEnd, runEnd)
    Plugin->>Core: stopBenchmark()
    Plugin->>Core: setExecutedBenchmark(pid, uri)
    Core-->>Runner: "FIFO: SampleStart > BenchmarkStart > BenchmarkEnd > SampleEnd"

    Note over Plugin,Runner: Bug in shared.ts (tinybench plugin)
    Plugin->>Core: startBenchmark()
    Plugin->>Core: currentTimestamp() → runStart
    Plugin->>Plugin: execute benchmark fn()
    Plugin->>Core: currentTimestamp() → runEnd
    Plugin->>Core: stopBenchmark() closes sample window FIRST
    Plugin->>Core: setExecutedBenchmark(pid, uri)
    Plugin->>Core: addMarker(BenchmarkStart, runStart) outside window
    Plugin->>Core: addMarker(BenchmarkEnd, runEnd) outside window
    Core-->>Runner: Markers outside sample - flame graph attribution breaks

Loading

_{Reviews (20): Last reviewed commit: "fixup: stop before sending timestamps" | Re-trigger Greptile}

[GuillaumeLagrange]

olgtm

[avalanche-staging]

Congrats! CodSpeed is installed 🎉

🆕 1 new benchmark was detected.

You will start to see performance impacts in the reports once the benchmarks are run from your default branch.

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

Detected benchmark

• fibo darwin (WallTime): 7.8 ms

---

[GuillaumeLagrange]

lgtm

[GuillaumeLagrange]

Although CI is failing.
Some of it is due to github instabilities of yesterday, but one of the job failed typechecking, I think because of core library. I'll elt you check