Control: severity -1 important Hi Uli,
On Sun, Sep 19, 2021 at 10:22:43AM +0200, Uli Schlachter wrote: > I took a look at run.sh (the script that integration.sh actually runs) > and as far as I can tell, this script exited successfully. > > The end of [0] is: > > autopkgtest [19:16:07]: @@@@@@@@@@@@@@@@@@@@ summary > integration.sh FAIL stderr: The XKEYBOARD keymap compiler > (xkbcomp) reports: > > So... is the autopkgtest failing because xkbcomp prints something on > stderr...? Not because something actually failed...? Yes, that is correct. The autopkgtest fails because it does not allow output on stderr, and recently xkbcomp started printing these errors. I don't fully understand why this is the case, it might also be related to some other x11-related library (in a previous similar case, also #953032, a rebuild of libx11 fixed it. Something similar is suggested in a upstream bug [0]). > If so, I wonder how this test ever passed. I don't think I ever xkbcomp > being quiet and not complaining about something. In the past it was actually the case that the whole test was stderr-clean (see old logs in [1]). :-) > Would "just pipe the output of xkbcomp to /dev/null" be a valid fix then? This problem could be fixed by either: 1) allowing stderr in the test run (adding the allow-stderr restriction in debian/tests/control). 2) filtering only these temporarily expected errors from stderr 3) getting rid of the whole xkbcomp output as you suggested, as it's unrelated to awesome functionality. 4) waiting until xkbcomp is fixed. I don't want to do 1 and 2, as 1 could hide other problems in the future, and 2 is just a hack for a temporary problem. So far I tried waiting for xkbcomp (#953032) to get fixed, as awesome is still working fine, and the failure should resolve itself eventually. If you as upstream think that xkbcomp's output is never relevant for the test suite result, we could apply your suggestion. Though so far I couldn't find where/how xkbcomp is invoked. If you have an idea, I could upload a fix. Though I also don't mind waiting for a fix in xkbcomp. @Paul / Release Team: I'm lowering the severity, as the failure is not a regression of awesome, but a test regression caused by another package. Solutions for this regression are mentioned above. If you disagree with the lowered severity, please raise it again. I will then filter/drop xkbcomp's output. Kind regards, Reiner [0] https://gitlab.freedesktop.org/xorg/lib/libx11/-/issues/137 [1] https://ci.debian.net/packages/a/awesome/
signature.asc
Description: PGP signature
