Hi Guillem, thanks for your feedback! However I'm not sure what to do with it...
On Sun, Oct 14, 2018 at 01:20:22AM +0200, Guillem Jover wrote: > Just in case this is helpful in the future, I've reproduced this now, > by using in addition: [...] interesting, thanks for sharing! > Then, first, there's a bug (normal probably) in dpkg with the > assert/internal-error, which should be a proper error message, this > needs to be fixed, and I think there's one or two bugs about just that. > Then, the current reporting in the buster internal error message still > sucks, so I'll try to improve that too, to: > > - Print the current processed package and in the ones in the queue. > - Print the command-line options used to invoke dpkg, or at least > the state of the --[no-]triggers option, and whether we were > called with a --pending or an explicit list of packages. > - Recommend running «dpkg --audit» and attaching that too, while > reporting the problem. cool! > But in any case the real cause for this seem to be just the usual trigger > cycle problem (serious, for the offending cycle introducer), that gets > abandoned and resets the state of dbus, which is never tried to get > configure again, as specified. > > The trigger happens with this: hm. I'm surprised we dont find any trigger cycles using https://jenkins.debian.net/job/dpkg_buster_find_trigger_cycles/ nor https://jenkins.debian.net/job/dpkg_sid_find_trigger_cycles/ nor cc:ing Josch who wrote https://salsa.debian.org/qa/jenkins.debian.net/blob/master/bin/find_dpkg_trigger_cycles.sh (on which those jobs are based) for input. > And dbus as seen from «dpkg --audit» is half-configured, and was not > in the processing queue, so dpkg was unable to make progress. > > After the assert/internerr, if you just run «dpkg --configure --pending» > dpkg will be able to make normal progress. well, that's nice for actual users but not helpful for CI jobs :) So I guess what's currently still missing is a serious bug against the package causing the reported behaviour in the first place, probably by just cloning this bug and reassigning it. Just to where to reassign is still unclear to me. -- cheers, Holger ------------------------------------------------------------------------------- holger@(debian|reproducible-builds|layer-acht).org PGP fingerprint: B8BF 5413 7B09 D35C F026 FE9D 091A B856 069A AA1C
signature.asc
Description: PGP signature
