Martijn Visser created FLINK-41041:
--------------------------------------
Summary: Maven waits until the time budget after a JDK 8 surefire
fork crashes with SIGSEGV in libjvm.so
Key: FLINK-41041
URL: https://issues.apache.org/jira/browse/FLINK-41041
Project: Flink
Issue Type: Bug
Components: Build System / CI
Affects Versions: 1.20.6
Reporter: Martijn Visser
Twice on release-1.20 a surefire fork on the Ubuntu OpenJDK 8u502 build crashed
with the same SIGSEGV in libjvm.so, and Maven then waited without output until
the job was killed at the time budget, about three and a half hours later:
https://github.com/apache/flink/actions/runs/35808755660/job/107024582937
(nightly 2026-09-23, AdaptiveScheduler, module connect, flink-csv)
https://github.com/apache/flink/actions/runs/38837104766/job/116781773746 (push
2026-10-16, module misc, flink-protobuf)
{code}
# SIGSEGV (0xb) at pc=0x00007fd74ec6a150, pid=115137, tid=0x00007fd7337fd6c0
# JRE version: OpenJDK Runtime Environment (8.0_502-b07) (build
1.8.0_502-8u502-ga~us1-0ubuntu1~24.04-b07)
# Problematic frame:
# V [libjvm.so+0xa56150]
Aborted (core dumped)
...
Process produced no output for 900 seconds.
...
"main" #1 prio=5 os_prio=0 waiting on condition
at java.util.concurrent.FutureTask.get(FutureTask.java:191)
at
org.apache.maven.plugin.surefire.booterclient.ForkStarter.awaitResultsDone(ForkStarter.java:432)
at
org.apache.maven.plugin.surefire.booterclient.ForkStarter.runSuitesForkOnceMultiple(ForkStarter.java:358)
{code}
The crash happens right after a new fork starts, before it reports a test
class; the other forks finish their classes and only MavenWrapperMain is left
in jps. Both crashes have the same offset (libjvm.so+0xa56150) and the 8u502
JDK came with the Ubuntu 24.04 image from FLINK-40548. On Temurin 17 the fork
crashes we saw (exit 134) failed the build right away instead of hanging.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
