beignet (and possibly other arch-specific packages) trapped in -proposed

Rebecca N. Palmer rebecca_palmer at zoho.com
Sun Feb 24 09:43:41 UTC 2019


(Previously sent to ubuntu-devel at lists.ubuntu.com as suggested by 
https://wiki.ubuntu.com/ProposedMigration#How_could_installing_a_package_into_the_release_pocket_possibly_break_other_packages.3F 
, but still in moderation after 10 days.  If "non-developers are 
moderated" in practice means "non-developers can't post", should this 
wiki page and/or the list's description / autoreply message be changed?)

britney tries to run autopkgtests on all architectures, including ones 
where the package under test does not exist:
https://bugs.launchpad.net/britney/+bug/1815131

For most arch-specific packages this is likely to produce a harmless 
(other than wasted resources) "Always failed".

However, beignet's tests are all skippable, and were added before Ubuntu 
autopkgtest had the skippable Restriction or the "all tests skipped" 
neutral state.  Hence, autopkgtest thinks they used to pass on 
non-beignet architectures, when they never actually did:
http://autopkgtest.ubuntu.com/packages/beignet

As they currently fail as uninstallable, this is considered to be a 
regression trapping the package in -proposed:
http://people.canonical.com/~ubuntu-archive/proposed-migration/update_excuses.html#beignet

As it thinks they used to *pass*, marking the tests skip-not-installable 
probably wouldn't be enough on its own.  I could do that and also add a 
"pass if beignet doesn't exist here, skip if it does" test 
(Test-Command: true, Depends: @, Conflicts: @, Restrictions: 
skip-not-installable ? though I haven't actually tried that), but that 
feels silly, and would also make beignet skip rather than fail if it 
became uninstallable on an architecture where it does exist.

Please let beignet out of -proposed:
https://bugs.launchpad.net/ubuntu/+source/beignet/+bug/1815014
and consider checking whether any other packages are in a similar state.





More information about the Ubuntu-devel-discuss mailing list