| Summary: | gthumb fails to install configration files correctly | ||
|---|---|---|---|
| Product: | [Build System, Metadata & Runtime] OE-Core | Reporter: | Saul Wold <sgw> |
| Component: | core | Assignee: | Saul Wold <sgw> |
| Status: | RESOLVED FIXED | QA Contact: | |
| Severity: | normal | ||
| Priority: | Medium | CC: | dvhart, jessica.zhang, meta.mr.watcher, meta.watcher, ross.burton |
| Version: | 1.2 | ||
| Target Milestone: | 1.3 | ||
| Hardware: | x86 | ||
| OS: | Multiple | ||
| Whiteboard: | (P2) | ||
| OS type for building Yocto: | --- | Type of Regression: | --- |
| Verified: | Documentation change: | --- | |
|
Description
Saul Wold
2012-04-04 21:49:06 UTC
This isn't strictly a gthumb issue. Looks like gconf.bbclass needs some robustness for highly parallel builds. If someone else has time to look at this that would be appreciated. When do the messages reported in Comment 1 appear? Which log? (In reply to comment #2) > When do the messages reported in Comment 1 appear? Which log? gconftool is used during postinst I believe so they'd be in log.do_rootfs Turns out that gconf schema registration was happening in make install, which is wrong for a packaging system. poky-contrib/ross/gconf should go towards fixing this, and might even solve the parallel problem in gthumb[1] [1] http://git.yoctoproject.org/cgit.cgi/poky/commit/?id=688b1ca792a9ce5b3aff3fd3b0fa075fca5a8e81 My hunch is that this problem is caused by multiple recipes running their do_install stage at once and simultaneously running gconftool to register their schemas, so they're racing to badly do atomic writes. This shouldn't happen as schema registration should happen in postinst, package builds should be disabling the registration process. I've just sent a patch to oe-core to fix this. (moving to core as it's a problem in gconf.bbclass) Fixed gconf.bbclass merged. (http://git.yoctoproject.org/cgit.cgi/poky/commit/?id=d3343f7307290e14eb90226a510de350b1cfd8b2) |