[not-matthias]

In simulation mode, register the running Node.js executable (process.execPath) on callgrind's obj-skip list when setupCore runs.

The runner only passes --obj-skip for the node it resolves on PATH. When benchmarks run a different binary, for example one provided by a Bazel toolchain, that skip never matches and callgrind instruments the whole V8 runtime. The profile then holds thousands of native V8 functions between the JS frames: the flamegraph is drowned in runtime internals and becomes very expensive to process.

This uses the existing instrument_hooks_callgrind_add_obj_skip client request (instrument-hooks resolves the real path), the same mechanism pytest-codspeed uses to skip sys.executable. The call happens in setupCore, before any benchmark starts the instrumentation, so the skip applies to all code the benchmark runs. Outside Valgrind it is a no-op, so walltime and memory modes are unaffected.

Checked locally under valgrind-codspeed with a copy of node at a path that differs from the one passed to --obj-skip: without the call, callgrind recorded 2,655 functions of that executable (1.5 MB profile). With it, it recorded none (60 KB profile). I did not test vitest workers running under Bazel.

[greptile-apps]

Confidence Score: 4/5

[Medium impact] The change reaches Callgrind profiling across Node benchmark integrations.

The PR appears safe to merge, with a non-blocking request for regression coverage.

Findings

1.  Alternate executable lacks test coverage ▶

Fix with agent prompt

### Issue 1
packages/core/src/index.ts:32
No automated test covers the alternate executable this fixes. Add a regression test that runs a copied Node binary under callgrind while `--obj-skip` names a different binary, then checks that the running binary contributes no native functions. The manual check would not catch a later change that removes this call or moves it after instrumentation starts.

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

Summary

This PR excludes the running Node executable from callgrind during setupCore.

• Callgrind skips the Node.js executable that runs each benchmark.

Diagram

sequenceDiagram
    participant Core as setupCore
    participant Native as Native wrapper
    participant Hooks as instrument-hooks
    participant Callgrind
    Core->>Native: callgrindAddObjSkip(process.execPath)
    Native->>Hooks: instrument_hooks_callgrind_add_obj_skip(path)
    Hooks->>Hooks: Resolve real path
    Hooks->>Callgrind: Add object to skip list
    Note over Core,Callgrind: Runs before the benchmark starts instrumentation

Loading

_{Reviews (1) · Last reviewed commit: "fix(core): exclude the running node exec..." · Reviewed by Greptile}

[codspeed]

Merging this PR will degrade performance by 12.74%

3 warnings, 1 note

⚠️ Walltime results may be unreliable

Some benchmarks used the Walltime instrument outside CodSpeed Macro Runners, which can make the results inconsistent.

Learn more about Macro Runners

See affected benchmarks

⚠️ 14 benchmarks are dominated by system calls

System calls cannot be simulated consistently and are left out of the measure. The reported value understates the real cost. Switch to the Walltime instrument to measure them.

Learn more about system calls

See affected benchmarks

⚠️ Base and head ran on different runner types

Some benchmarks with significant performance changes were compared across different runner types, so the changes might not come from the code. Run base and head on the same runner type, or use CodSpeed Macro Runners for more stable results.

Learn more about Macro Runners

Open the report in CodSpeed to investigate

ℹ️ 16 benchmarks have no flamegraph

The flamegraph wasn't generated: most of the time was spent in runtime internals, which would leave nothing useful to analyze.

See affected benchmarks

⚡ 7 improved benchmarks
❌ 4 regressed benchmarks
✅ 253 untouched benchmarks
⏩ 56 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	CPU Simulation	switch 1	245.6 µs	597.9 µs	-58.92%
❌	CPU Simulation	recursive fibo 10	338.6 µs	692.6 µs	-51.12%
❌	CPU Simulation	recursive fibo 10	338.7 µs	688.9 µs	-50.83%
❌	Memory	wait 1sec	22 KB	28.1 KB	-21.44%
⚡	Memory	wait 500ms	28.2 KB	22.7 KB	+24.27%
⚡	CPU Simulation	wait 500ms	46.1 ms	38.7 ms	+19.21%
⚡	CPU Simulation	wait 500ms	45.3 ms	38.1 ms	+18.84%
⚡	CPU Simulation	wait 1sec	84.8 ms	72.9 ms	+16.27%
⚡	CPU Simulation	wait 1sec	83.2 ms	72.9 ms	+14.13%
⚡	Walltime	fibo darwin	9.2 ms	8.2 ms	+12%
⚡	Walltime	test_iterative_fibo_10	132 ns	120 ns	+10%

Tip

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

---

_{Comparing fix/obj-skip-running-node-executable (727b808) with main (5e52a1d)}

Footnotes

1. 56 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. ↩

[GuillaumeLagrange]

lgtm BUT i'm a bit skeptical about the introspected node thing if this is actually a real thing 🤔