Skip to content

[DNM][CI] Debug spark-test-spark41 SIGSEGV in NativeBackendInitializer#12552

Draft
jackylee-ch wants to merge 6 commits into
apache:mainfrom
jackylee-ch:debug-spark41-ci
Draft

[DNM][CI] Debug spark-test-spark41 SIGSEGV in NativeBackendInitializer#12552
jackylee-ch wants to merge 6 commits into
apache:mainfrom
jackylee-ch:debug-spark41-ci

Conversation

@jackylee-ch

Copy link
Copy Markdown
Contributor

What changes were proposed in this pull request?

DNM — Debug-only PR for issue #12534.

spark-test-spark41 / spark-test-spark41-slow fail consistently on every recent Velox Backend (x86) run. The forked surefire JVM dies with SIGSEGV (exit 134) while running ColumnarBatchTest (the first VeloxBackendTestBase-based suite), inside native backend initialization:

siginfo: si_signo: 11 (SIGSEGV), si_code: 1 (SEGV_MAPERR), si_addr: 0x0000000000000030
C  [libgluten.so+0xdb73a4]
j  org.apache.gluten.init.NativeBackendInitializer.initialize(...)
j  org.apache.gluten.backendsapi.velox.VeloxListenerApi.initialize(...)
j  org.apache.gluten.test.VeloxBackendTestBase.setup()

(hs_err captured via the diagnostics steps borrowed from #12551's run.)

This PR trims velox_backend_x86.yml to only build-native-lib-centos-7 + the two spark41 jobs, and adds crash-diagnostics steps (print/upload hs_err_pid*.log, surefire *.dumpstream, cores) so we can iterate quickly and bisect the cause.

How was this patch tested?

CI only.

@github-actions github-actions Bot added the INFRA label Jul 18, 2026
@github-actions github-actions Bot added the VELOX label Jul 18, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant