Hi Maximiliano, others On Mon, 3 Dec 2018 11:24:15 -0300 Maximiliano Curia <maxy@gnuservers.com.ar> wrote: > About the kde frameworks uploads, they are handled in a bundle, and breaks and > dependencies are added so they migrate to testing as needed, the same breaks > and dependencies can cause temporary uninstallability in unstable, thus, > having a lot of temporary regressions, in order to have a "smoother" testing. > > At the same time, the autopkgtest run (mostly) upstream's unittests (mostly > the ones that can't be run as part of the build). But, sadly, upstream doesn't > enforce running their unittests as part of their development nor release > process, so, some regressions in part of their unittests is "normal". This > fits "nicely" with the current way Debian delays the transitions to testing, > which gives enough time to unstable users to report real regressions in > behaviour. > > All of this is to explain that, currently, a report about autopkgtest issues > is not really useful to us, and would be time better spent reporting them > directly upstream (if it applies). Furthermore, once the autopkgtest are used > to block the migration to testing completely, we would be forced to to simply > drop most of the tests (as long as upstream doesn't change their policy > regarding unittest). > > This can be considered as a documentation of the current state of affairs > regarding kde frameworks and kde plasma. No response needed. However, currently regressions due to glibc in kio, kdelibs4support, khtml and kparts are blocking the migration of glibc to testing. Should we be ignoring those regressions as well? If that is the case, can you please already start removing the tests, as they also hamper other packages, or maybe you want to mark them as "flaky" [1]. All four fail on the abi-compliance test. Paul [1] https://salsa.debian.org/ci-team/autopkgtest/raw/master/doc/README.package-tests.rst
Attachment:
signature.asc
Description: OpenPGP digital signature