Control: retitle -1 igdiscover: test suite flaky, possibly on high core count machines Control: severity -1 important Control: tag -1 confirmed
Hi Christian, Hi Nilesh, Christian Kastner, on 2021-02-11 15:46:16 +0100: > On 11.02.21 15:20, Nilesh Patra wrote: > > According to last reproducible build result[1] shows that it is building > > successfully on January 31, 2021 (~52 days after this bug was reported) > > However there are a few FTBFS entries in there as well. So probably this is > > an occasional failure? > > > > [1]: > > https://tests.reproducible-builds.org/debian/history/amd64/igdiscover.html > > I tried N=1 build, and this time it succeeded. > > But Nilesh is right, reproducible builds failed occasionally, with the > errors as I reported above. It looks as if some of the tests are flaky. Thanks for your observations. The test suite stubbornly succeeds on my six cores equipment. I see failures on amd64 and arm64 machines in reproducible builds reports. If I'm right, those hosts may have notoriously high cores count (>100). The issue might be caused by a race to read/write test data, or something more subtle. I would guess a path of least resistance would be to drop parallelization, at least for the test suite. I'm afraid I'm not having an appropriate configuration to wrap up a fix and test that kind of issues at the moment. > Now, because these failures are not a pytest issue, I don't think this > bug is valid any longer -- pytest 6 compatibility was the goal, and it > is present. > > Please feel free to downgrade and retitle for the flakiness issue. Thanks, the bug meta-informations should be accurate now, modulo typos. Have a nice day, :) -- Étienne Mollier <etienne.moll...@mailoo.org> Fingerprint: 8f91 b227 c7d6 f2b1 948c 8236 793c f67e 8f0d 11da Sent from /dev/pts/2, please excuse my verbosity.
signature.asc
Description: PGP signature