Hi,
I've just noticed that there is a race condition in log_check after rootfs
creation. do_rootfs task invokes package type specific shell function - in my
case it's rootfs_ipk_do_rootfs which in turn invokes log_check on the log file
produced by do_rootfs (log.do_rootfs.${PID}), which is buffered by BitBake, and
despite the fact that all packages are already installed when log_check runs,
output may not have been flushed yet, so some errors may not be detected.
In my case relevant parts of log.do_rootfs look like that:
#v+
Collected errors:
* opkg_conf_load: Could not create lock file [...]
#v-
And a few lines below:
#v+
log_check: Using [...]
Logfile is clean
#v-
and it was consistently passing log_check despite this error being there for a
long time. I'm using slightly outdated OE, but there haven't been any relevant
changes since then.
Regards,
Zbigniew
_______________________________________________
Openembedded-core mailing list
[email protected]
http://lists.openembedded.org/mailman/listinfo/openembedded-core
