Ludovic Courtès <[email protected]> writes: > Hi Clément, > > Clément Lassieur <[email protected]> skribis: > >> For some time after your message, only 8 of those 16 tests seemed to >> fail. But now they all fail again[1]. > > These GuixSD installation tests occasionally fail for several reasons: > > • Sometimes I mess up with ‘guix system’ and the tests catch the > regression (e.g., commit a29ce1f97dc8336e5a063f14d98d673b33b4bff4). > > • Sometimes there are missing files in the Guix tarball, which breaks > ‘guix’ usage in the installation VM (commit > 1e5b79e5d0e6629dce838dbb10febad676eea773). > > • Sometimes installing bare-bones takes slightly more than 1G and the > tests assume 1G is enough (which is a good thing IMO, because it > ensures the image size doesn’t grow unreasonably.) > > On this last point I expect we’ll do better when ‘core-updates’ is > merged because currently on ‘master’ there’s both Guile 2.2.4 and 2.2.3 > in the system closure (the latter coming from the ‘guix’ package). > > Overall I think these tests are doing a good job a finding problems. :-) > > The problem is that these issues lead to dependency-failed failures on > Cuirass/Hydra, and Cuirass currently doesn’t allow you to see which > dependency failed and to view its build log (I can view that by logging > in on berlin but that’s obviously suboptimal.) We should fix Cuirass to > display that info though I’m not entirely sure how.
https://bugs.gnu.org/32954 I guess. > Thoughts? Thank you for the explanation. So they are different failures. It's probably not worth keeping a bug open for them, so I'm closing it again. Cheers, Clément
