Preserve Python and Node symbols in memory flamegraphs

[not-matthias]

Summary

• Capture per-process perf maps and Python JIT unwind data alongside native memory artifacts.
• Enable V8 perf maps for memory runs while preserving existing NODE_OPTIONS.
• Document native-allocation coverage and runtime limitations.
• One weird edge case where binaries can be mapped twice into a single address space

Verification

• cargo test --release --bin codspeed writes_keyed_artifacts_and_metadata_for_a_streamed_mapping
• cargo fmt --all --check
• Local Python 3.12 perf-trampoline and Node 22 memory runs produced perf maps and resolved language frames when parsed offline.

[codspeed]

Merging this PR will degrade performance by 10.47%

⚠️ 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.

❌ 2 regressed benchmarks
✅ 29 untouched benchmarks
⏩ 6 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	WallTime	memtrack track tar	8.4 s	9.4 s	-10.71%
❌	Memory	encode_events_realistic[8]	169.6 MB	189 MB	-10.24%

Tip

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

---

_{Comparing cod-3654-support-memory-flamegraphs-for-pythonnode (a01c776) with main (214c040)}

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

[High risk] Adds Python and Node.js symbol support to memory profiling.

The PR is not yet safe to merge because plain Node wall-time benchmarks now run with a memory-specific V8 option that can alter their measurements.

Findings

1.  Wall-time runs gain profiling overhead ▶
2.  JIT unwind records get discarded ▶
3.  Node symbol support overstated ▶

Fix with agent prompt

### Issue 1
crates/exec-harness/src/node.rs:5
For a plain Node.js wall-time benchmark, this shared option list now adds `--interpreted-frames-native-stack` to every timed child; previously it was memory-only. The flag changes how V8 runs interpreted frames and increases its code footprint, so measured times can include profiling overhead unrelated to the workload. Keep this option specific to memory runs.

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!

### Issue 2
src/executor/memory/module_artifacts.rs:65-70
If appending one process's JIT symbols fails after another process was processed, this fallback discards the successful process's unwind records too, while leaving its symbols in place. The completed profile may name those frames but be unable to unwind through them. Preserve records collected for processes that succeeded.

### Issue 3
README.md:167-168
Removing the `codspeed exec` qualification implies that Node.js memory flamegraphs have JavaScript function names through any supported integration. Exec-harness enables perf maps, but the direct [Node integration's memory-mode flags](https://github.com/codspeedhq/codspeed-node/blob/HEAD/packages/core/src/introspection.ts) omit `--perf-basic-prof`. Users of that path can therefore get native allocations without the JavaScript function names this PR aims to preserve. Retain the qualification or document the required runtime configuration.

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

The PR collects runtime perf maps and Python JIT unwind data for memory profiles, retains multiple placements of a mapped native module, and adds interpreter coverage tests.

• The latest changes make harvest failures nonfatal, extend a Node profiling option to all exec-harness modes, add the interpreter tests to CI, and revise the memory documentation.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  B[Benchmark process] --> M[Memtrack mappings and stacks]
  B --> P[Runtime perf maps and JIT dumps]
  M --> A[Native symbol and unwind artifacts]
  P --> A
  A --> D[memtrack.metadata and profile folder]

Loading

_{Reviews (3) · Last reviewed commit: "fix(memory): keep every placement of a m..."}

[GuillaumeLagrange]

olgtm, second round will be quick