[goutamadwant]

Summary

Surefire 3.6.0 uses StackWalker for stack capture on Java 9+. Issue #3459 reports that three RocksDB FFM callback test classes terminate their forked JVMs on Windows x64 with JDK 25. The downstream workaround sets stackTraceMaxFrames to 0, which also removes stack capture.

This change adds the JVM property surefire.stackTrace.disableStackWalker=true to select the existing Thread.getStackTrace() path while retaining stack capture. StackWalker remains the default. For forked tests, the property can be appended to argLine; the documentation includes an example. Regression tests cover the default and opt-out paths.

Related to #3459.

Validation

• mvn clean install passed locally on JDK 21.
• Focused tests passed on JDK 8, and the affected console-output integration tests passed locally.
• mvn clean install passed in a Linux arm64 JDK 25 container.
• A Windows x64/JDK 25 run of the three named classes passed with StackWalker enabled. This narrower run did not reproduce 3.6.0 regression: forked JVM silently vanishes on Windows for tests using FFM upcalls from background threads #3459.
• Full-suite Windows x64/JDK 25 validation against Surefire 3.6.0 and the patched opt-out is pending: https://github.com/goutamadwant/maven-surefire/actions/runs/36351047252
• The fork's Verify matrix passed 14 of 15 jobs. The Linux JDK 8 job failed in the unchanged ProcessCheckerTest.shouldHavePidAtBegin PID assertion: https://github.com/goutamadwant/maven-surefire/actions/runs/36345975232

Contribution checklist

• The commit has a meaningful subject and body.
• This description explains the change and its purpose.
• mvn clean install passed.
• mvn -Prun-its clean install passed.
• I hereby declare this contribution to be licenced under the Apache License Version 2.0, January 2004.
• If required, I have filed an Apache Individual Contributor License Agreement.

[olamy]

sure we can add some workaround but the root cause seems to be a JDK bug though

[github-actions]

@olamy Please assign appropriate label to PR according to the type of change.