[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Bug#945113: transition: hwloc

Hi Samuel,

On 28-11-2019 09:24, Samuel Thibault wrote:
> https://tracker.debian.org/pkg/hwloc
> https://ci.debian.net/data/autopkgtest/testing/amd64/libr/librsb/3524853/log.gz
> shows that autopkgtest fails in testing for librsb, thus blocking
> the libhwloc.so.5 -> libhwloc.so.15 transition, but that's actually
> _because_ of the transition:
> [...]
> Get:7 http://deb.debian.org/debian testing/main amd64 librsb0 amd64 [687 kB]
> [...]
> compile-examples     FAIL stderr: /usr/bin/ld: warning: libhwloc.so.5, needed by /usr/lib/x86_64-linux-gnu/librsb.so, may conflict with libhwloc.so.15
> i.e. it installed the non-rebuilt librsb0 from testing
> and then noticed the conflict warning.  If it was installing the rebuilt
> librsb from unstable (which will migrate to testing
> along hwloc) it would not get the warning and thus succeed.
> I don't know if such a case is known by the release and CI team, just
> making sure here.

Indeed, the situation during transitions isn't ideal. Regularly the
issue is a lot worse until all relevant packages are rebuild, and than
the retries will sort the situation out. Also, in case of transitions,
we indeed don't want to test against the packages in testing that link
to the old library, but against the rebuilds. I have tried already a
couple of times to start on a fix in britney, but so far I haven't
figured out what we exactly want, let alone implement it. The trouble is
that by doing "the right thing" we basically loose smooth upgrades, as
all reverse dependencies need to be successfully rebuild in that case.

So, I don't think the situation is very bad at this moment. In this case
the test doesn't fail by itself, the reason why the autopkgtest fails is
due to a WARNING on stderr. The autopkgtest of librsb follows the
default behavior of failing on output to strerr, which, in cases like
this shows that could be the wrong setting for this package if the test
is otherwise showing that the package works. Adding the maintainers into
the conversation to get their opinion.


Attachment: signature.asc
Description: OpenPGP digital signature

Reply to: