refactor: spawn child processes with tokio::process

[lvaroqui]

Spawn the benchmark, build and MongoDB tracer processes with tokio::process instead of std::process, so waiting on a child no longer blocks the runtime.

RunnerFifo::handle_fifo_messages now selects on incoming FIFO commands and on child.wait(). Previously it only checked try_wait() after recv_cmd had been idle for 1s, so each run returned up to a second after the benchmark process exited.

Commands still unread when the child exits can be dropped, as before: the old loop could also time out, receive a command and see the exit in try_wait() before reading it. Now the loop returns as soon as child.wait() resolves, and if a command and the exit are ready at the same time select! picks one at random. What actually prevents lost commands is the ack: integrations send each command and block until the runner acks it, so they can't exit with a command unread.

Child stdout and stderr are taken back as blocking files (take_blocking_stdio, built on into_owned_fd, which clears O_NONBLOCK) and drained on blocking threads, as before.

Closes COD-3863

[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.

⚠️ 8 benchmarks spent significant time in system calls

System calls cannot be consistently instrumented, so they are not included in the measure, which understates the real cost. Please switch to the Walltime instrument to accurately measure system calls.

Measurement and system calls

• encode_events_realistic[8]
• encode_events_realistic[16]
• encode_events_realistic[4]
• find_module_events[1000000]
• find_module_events_with_stacks[1000000]
• encode_events_realistic[2]
• sleep 1
• encode_events_realistic[1]

✅ 41 untouched benchmarks
⏩ 6 skipped benchmarks¹

---

_{Comparing cod-3863-chore-avoid-blocking-in-runner (2c52924) with main (9e74317)}

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

[High impact] The PR appears safe to merge.

Summary

Moves benchmark, build, and MongoDB tracer processes to tokio::process.

• The runner returns profiling results as soon as the benchmark child exits.
• Benchmark, build, and tracer child processes use Tokio while logs stay on blocking readers.

No actionable issues were found.

lvaroqui explicitly identifies dropping unread FIFO commands at child exit as intentional. The stated safeguard is that integrations wait for a reply before continuing.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  A[Spawn child with Tokio] --> B[Drain output on blocking threads]
  A --> C[Wait for FIFO command or child exit]
  C -->|Command| D[Handle command and send reply]
  D --> C
  C -->|Child exits| E[Return timestamps and exit status]
  B --> F[Finish draining output]
  E --> F
  F --> G[Return result]

Loading

_{Reviews (1) · Last reviewed commit: "refactor: spawn child processes with tok..." · Reviewed by Greptile}

[not-matthias]

Nice job!

[not-matthias]

We must ensure that we don't accidentally discard events inside the FIFO when the process exits. I suggest we add biased; to the recv_cmd branch to ensure we always finish reading the FIFO first (see: https://docs.rs/tokio/latest/tokio/macro.select.html#fairness)

tokio::select! {
+	// Read commands already in the FIFO before acting on the exit.
+	biased;
	result = self.recv_cmd() => {

[not-matthias]

Nice job!