Hi all, On Sun, Jun 16, 2019 at 10:59:05AM +0300, Niko Tyni wrote: > Package: reprotest > Version: 0.7.8 > > My package has a test which don't work under faketime (see #930441), > so I'm trying to eliminate that from the build command in reprotest but > had no success so far. > > It looks like reprotest is using faketime to fix the time of the first > build regardless of the variations listed.
I am in the same situation with my “sphinx” package. Its pipelines started failing on GitLab, see for example [1]. I noticed that the control build has failed, not (just) the build with actual variations. The tests that failed are sensitive to timestamps: they are testing the function that returns number of files needing rebuild. One of the tests creates a file and then asserts that it does not need rebuild (its mtime is less that the current time). However, as reprotest calls faketime with NO_FAKE_STAT=1, the file is created with the real time while the clock is a bit behind (I guess it is set to the result of auto_source_date_epoch() function). So the file mtime is in future compared to the clock, which makes the test fail. So far I have found no workaround to this issue. [1]: https://salsa.debian.org/python-team/modules/sphinx/-/jobs/221195 -- Dmitry Shachnev
signature.asc
Description: PGP signature
