feat: add setTraceId and clearTraceId to ProfilerApi

[marcsanmi]

Summary

Adds two methods to ProfilerApi so the OTel integration can attach the trace ID as a string label on profile samples. setTracingContext(long, long) cannot carry a 128 bit value, hence a separate entry point.

The ProfilerSdk impl backs them with a ThreadLocal<ScopedContext> that opens ScopedContext("trace_id", traceId) on set and closes it on clear, reusing the existing labels pipeline.

Both new methods are declared default (no op) on the interface so older runtimes stay wire compatible with newer callers (the label simply does not appear).

Related PR https://github.com/grafana/otel-profiling-java/pull/TBD will land once this is released.

[cla-assistant]

All committers have signed the CLA.

[cla-assistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[korniltsev-grafanista]

I would prefer if we did not use the labels infrastructure (ScopedContext, LabelsSet) for this and instead passed trace id through setTracingContext and JFR (that would require async-profiler changes, we can put trace id as well as spanId and spanName into the thread local structure they recently added asprof_thread_local_data ).

The main reason is that the current implementation will break in few ways:

• If one already associated labels to the thread with ScopedContext, the new set of methods will not be aware of it and silently override it.
• Similarly the other way around: if we set labels with setTraceId and then someone else uses ScopedContext - it will override trace id label, unless we expose the new java ThreadLocal somehow, which may be not trivial, especially across class loaders.

Another reason is performance (premature?) optimizations. I got no proof, but gut feeling tells me storing __int128 would be cheaper then keeping every traceId strings between profile uploads.

[korniltsev-grafanista]

String#substring allocates, can we somehow try parsing hexes without it?

[Copilot]

Pull request overview

This PR extends the cross-classloader ProfilerApi with setTraceId(String) / clearTraceId() so the OTel integration can associate a 128-bit W3C trace ID with samples, and updates the embedded async-profiler fork version/distribution artifacts accordingly.

Changes:

• Add setTraceId(String) and clearTraceId() as default methods on ProfilerApi for backward-compatible implementations.
• Implement setTraceId/clearTraceId in ProfilerSdk by parsing a 32-hex-char trace ID into two longs and delegating to async-profiler.
• Bump bundled Grafana async-profiler version and update .so.sha1 checksums; add unit tests for hex parsing.

Reviewed changes

Copilot reviewed 7 out of 11 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
Makefile	Bumps the Grafana async-profiler version used by the download script.
async-profiler-grafana-fork-dist/lib/libasyncProfiler-macos.so.sha1	Updates checksum for the macOS native library.
async-profiler-grafana-fork-dist/lib/libasyncProfiler-linux-x64.so.sha1	Updates checksum for the Linux x64 native library.
async-profiler-grafana-fork-dist/lib/libasyncProfiler-linux-arm64.so.sha1	Updates checksum for the Linux arm64 native library.
agent/src/test/java/io/pyroscope/javaagent/ProfilerSdkTest.java	Adds unit tests validating parseHex64 parsing behavior.
agent/src/main/java/io/pyroscope/javaagent/ProfilerSdk.java	Implements trace ID support and introduces parseHex64.
agent/src/main/java/io/pyroscope/javaagent/api/ProfilerApi.java	Adds the new API methods as defaults for compatibility.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.