The RPM signing test in oe-selftest sometimes fails because of a failure in tempfile creation. This is a very rare occurrence and only seen in autobuilder, so far. The interesting build log snippet is: ERROR: ed-1.9-r0 do_package_write_rpm: Function failed: Failed to sign RPM packages: error: rpmTempFile failed It would seem to indicate some kind of race condition inside rpmTempFile.
If the rpm log file for the failure were available, it might provide additional clues as to what is happening, as in some cases the error message there includes the full pathname of the temp file it is trying to create. It appears that the rpm log ends up as part of the log.do_package_write_rpm file.
It shows all rpm output in the bitbake output. There's no other rpm log, same output will be shown in log.do_package_write_rpm. But you have a good point in that no other error message is shown which narrows the possible points of failure. I've been trying to reproduce this without any success. Probably the way to try to solve is to add more (and more verbose) error messages and try to determine the cause from the log of next failed oe-selftest.
A patch for more detailed error messages was merged in http://git.openembedded.org/openembedded-core/commit/?id=aba432ea3ada509b24075c159081e8f0f942c308 This will hopefully give more details when it fails the next time. I suspect that the tmpdir initialization somehow fails...
AFAIK, we haven't seen this problem since April. So it seems to be VERY hard to reproduce. Closing it for now. Please re-open if we see the problem, again.
Reopening as we saw the problem, again
Patch sent for review: https://patchwork.openembedded.org/patch/132925/
Fix merged in master (and morty): http://git.openembedded.org/openembedded-core/commit/?id=84de3283fa2a2908d367eb58953903ae685b0298