jackylee-ch opened a new pull request, #12557:
URL: https://github.com/apache/gluten/pull/12557
## What changes were proposed in this pull request?
The `spark-test-spark41` / `spark-test-spark41-slow` jobs in `Velox Backend
(x86)` have been failing **consistently** (exit 134 / SIGABRT) since the
mid-July GitHub-hosted runner image rollout (tracked in #12534). The forked
Surefire JVM crashes with SIGSEGV during Velox backend initialization, before
the first native test (`ColumnarBatchTest`) even runs:
```
# C [libgluten.so+0xdb73a4]
std::__codecvt_utf16_base<char16_t>::do_unshift(...)+0x4
# siginfo: si_code: SEGV_MAPERR, si_addr: 0x0000000000000030
Java frames:
org.apache.gluten.init.NativeBackendInitializer.initialize(...)
org.apache.gluten.backendsapi.velox.VeloxListenerApi.onExecutorStart(...)
org.apache.gluten.test.VeloxBackendTestBase.setup()
```
`addr2line` against the CI native artifact resolves the faulting frame to
the **statically-linked libstdc++ (gcc 13.2) locale/codecvt facet** path,
reached via `std::ostream::_M_insert<long>`.
### Root cause
The crash only reproduces inside the `apache/gluten:centos-9-jdk17`
container (glibc 2.34 + system `libstdc++.so.6.0.29`), in combination with the
centos-7/gcc13 statically-linked native libraries and the new runner image.
Under the same native artifact, `spark-test-spark33` … `spark-test-spark40`
(which run in the `centos-8` image) stay green — so this is not a code
regression from any recently merged PR, but an environment issue specific to
the `centos-9` container image.
### Fix
Move the two Spark 4.1 jobs back to `apache/gluten:centos-8-jdk17` — the
same image used by the (green) Spark 4.0 jobs. Because CentOS 8 does not
package Python >= 3.10 (required by the Spark 4.1 python UDF tests), Python
3.11 is installed via miniconda in the `Prepare` step.
## How was this patch tested?
Validated on a debug branch that trims the workflow to only the affected
jobs. With the container switched to `centos-8-jdk17` + miniconda Python 3.11:
- `spark-test-spark41` — **success** (the crash is gone; `arrow batched
python udf over parquet scan` and the full backends-velox / gluten-ut suite
pass)
- `spark-test-spark41-slow` — **success**
Run: https://github.com/jackylee-ch/gluten/actions/runs/29644190820
Closes #12534
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
