fix(runner-shared): always set PYTHONPERFSUPPORT

[not-matthias]

TLDR: Since exec-harness 1.4.0, Python exec benchmarks in simulation mode have no Python frames in their callgraphs. The runtime env now sets PYTHONPERFSUPPORT=1 in every mode.

• The runtime env only set PYTHONPERFSUPPORT=1 in memory mode. Without it, Python writes no /tmp/perf-<pid>.map, so valgrind can't name Python frames. exec-harness 1.3.0 set it per command in simulation mode.
• Walltime on Linux is unchanged: PYTHON_PERF_JIT_SUPPORT=1 takes precedence in CPython, so jitdump is still used.

Review notes

• Simulation now sets the variable for every integration, not only exec-harness. pytest-codspeed already activates the perf trampoline itself.
• macOS walltime keeps jitdump off (COD-2645), so Python there now emits a perf map under samply. Not tested on macOS.
• The trampoline costs instructions, so Python exec simulation values go back up. Python 3.12 running a fractal script under callgrind: 2.22e9 Ir without it, 2.51e9 Ir with it (+13%).

Closes COD-3820

[codspeed]

Merging this PR will not alter performance

⚠️ 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.

✅ 31 untouched benchmarks
⏩ 6 skipped benchmarks¹

---

_{Comparing cod-3820-investigate-missing-python-valgrind-frames-in-runner-541 (aecc796) with main (24f182e)}

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: 5/5

[Medium impact] The PR appears safe to merge; no new actionable defect was found.

Summary

Sets PYTHONPERFSUPPORT=1 in every measurement mode so Python exec simulation benchmarks can provide frame names.

• Keeps the existing platform-dependent PYTHON_PERF_JIT_SUPPORT setting.
• Current pytest-codspeed skips activation when the trampoline is already active.
• not-matthias acknowledges the higher simulation instruction counts and the macOS walltime change, which has not been tested there.
• The previous test suggestion was manually resolved without explanation. It was unnumbered, so it has no previousFindings entry.

_{Reviews (2) · Last reviewed commit: "fix(runner-shared): always set PYTHONPER..." · Reviewed by Greptile}

[GuillaumeLagrange]

I could have caught this during review sorry

[not-matthias]

I could have caught this during review sorry

Also completely missed it. I'm a bit surprised Greptile didn't catch this too 🙃