[not-matthias]

TLDR: Reuse perf’s runtime unwinding inference in samply so detected Go and Java entrypoint commands use frame pointers (Linux only).

• Share the existing resolver; explicit unwinding settings take precedence.
• Python uses dwarf,32768 with both profilers. Other commands retain backend defaults: 32000 bytes for samply, 8192 for perf.
• Exec-harness target inference is unchanged; this does not add automatic detection inside serialized exec targets.
• Verified 299 library tests, Clippy, build, and live Java command launches selecting FP with both profilers. Smoke runs contained no instrumented benchmarks.

Stacked on #586; review against its branch.

Fixes COD-3861

[codspeed]

Merging this PR will not alter performance

2 warnings

⚠️ 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

⚠️ 8 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

✅ 41 untouched benchmarks
⏩ 6 skipped benchmarks¹

---

_{Comparing cod-3861-support-samply-fp-unwinding (3bd4427) with cod-3877-expose-stack-sampling-size-in-cli (f77fdc8)}

Footnotes

1. 6 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. ↩

[greptile-apps]

Confidence Score: 4/5

[Medium impact] The change affects Linux wall-time profiling with perf and samply.

Fix exec argument detection before merging so ordinary arguments cannot select the wrong unwinding mode.

Findings

1.  Arguments select the wrong unwinding ▶

Fix with agent prompt

### Issue 1
src/executor/orchestrator.rs:153-157
Every exec argument is checked as an executable name. For example, `codspeed exec -m walltime -- grep go corpus.txt` selects frame-pointer unwinding because the pattern `go` reaches `suggested_unwinding` as a command. If the native program was built without frame pointers, its call stacks will be incomplete. Every other target in the same recording also gets this setting.

Keep each target's argument list intact and distinguish its executable from ordinary arguments before inferring the runtime.

---

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

Summary

Moves unwinding inference to the orchestrator so it sees original benchmark targets and applies to both profilers.

• Linux wall-time runs choose unwinding from the original benchmark commands.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Original benchmark targets] --> B[Infer unwinding]
    C[Explicit setting] --> D[Resolve per-run setting]
    B --> D
    D --> E[Build benchmark environment]
    E --> F{Profiler}
    F --> G[perf call graph option]
    F --> H[samply unwinding options]

Loading

_{Reviews (1) · Last reviewed commit: "fixup! fix(walltime): infer Go and Java ..." · Reviewed by Greptile}

[GuillaumeLagrange]

lgtm