| Summary: | Worker process exited unexpectedly during selftest | ||
|---|---|---|---|
| Product: | [Build System, Metadata & Runtime] OE-Core | Reporter: | Chee Yang <chee.yang.lee> |
| Component: | core | Assignee: | Richard Purdie <richard.purdie> |
| Status: | RESOLVED OBSOLETE | QA Contact: | |
| Severity: | normal | ||
| Priority: | Medium | CC: | alexandre.belloni, luca.ceresoli, meta.mr.watcher, meta.watcher, randy.macleod, ross.burton |
| Version: | unspecified | ||
| Target Milestone: | 5.0 | ||
| Hardware: | x86 | ||
| OS: | Multiple | ||
| Whiteboard: | AB-INT AB-INT-TMPFS AB-HAVE-PLAN | ||
| OS type for building Yocto: | --- | Type of Regression: | --- |
| Verified: | Documentation change: | Don't know | |
|
Description
Chee Yang
2021-05-12 02:36:26 UTC
https://autobuilder.yoctoproject.org/typhoon/#/builders/108/builds/1644/steps/12/logs/stdio qemux86-world-alt centos7-ty-4 Second failure is exit -9 - EBADF. First failure is exit 0. Maybe a race in the shutdown code somehow? I think the first failure was some kind of fault in bitbake-worker. The real failure was probably in the bitbake-cookerdaemon.log but long gone now. I've added a fix to bitbake to ensure we don't see the "exit 0" success code if there is a fault: http://git.yoctoproject.org/cgit.cgi/poky/commit/?id=4437754142f6d0e8f3cf9ffbb51c61321b8c5600 I think the second failure was the OOM killer in action with SIGKILL being the sig 9. Not much to be done about that. Closing, if we see this again we need the bitbake-cookerdaemon log to get any further and hopefully we won't see the exit 0 again. https://autobuilder.yoctoproject.org/typhoon/#/builders/86/builds/3988/steps/11/logs/stdio oe-selftest-fedora on fedora36-ty-3 bitbake-cookerdaemon.log and dmesg did not seem to contain useful clues. At least this time upon recurrence it did show exit 1 instead of 0 so we have some idea which code paths were triggering, but still no cooker log :( I saw this earlier in the week, so it's not disappeared... Likely still a problem so keeping it opened even though it hasn't happened in months. We've not logged this in a long time, closing and can reopen if it reappears. Assuming there were other fixes which stopped it happening. |