Hi James, On Sat, Nov 10, 2018 at 09:29:42AM -0500, James McCoy wrote: > On Fri, Oct 26, 2018 at 03:30:54PM +0200, Helmut Grohne wrote: > > neovim fails to build from source in unstable. > > > > | [ ERROR ] 6 errors, listed below: > > | [ ERROR ] test/functional/helpers.lua @ 743: after_each > > | test/helpers.lua:289: crash detected (see above) > > | > > | stack traceback: > > | test/helpers.lua:289: in function 'check_cores' > > | test/functional/helpers.lua:748: in function > > <test/functional/helpers.lua:743> > > I haven't seen these on the buidds or locally. The only issue I've been > seeing, aside from some flaky tests, I've filed upstream[0]: > > [ RUN ] mbyte utf_char2bytes: FAIL > test/unit/helpers.lua:731: (string) ' > not enough memory' > > stack traceback: > test/unit/helpers.lua:747: in function 'itp_parent' > test/unit/helpers.lua:774: in function <test/unit/helpers.lua:764> > > [0]: https://github.com/neovim/neovim/issues/9212
Thank you for looking into this. I've no clue what the error is supposed to mean either. In any case, I just retried and I can still reliably reproduce the failure. > > Looking to reproducible builds, the failure pattern looks quite random: > > > > https://tests.reproducible-builds.org/debian/rb-pkg/unstable/amd64/neovim.html > > I'm not actually sure what's happening in the most recent builds there. > I don't see any explicit error output, just the testing stopping earlier > than it should. That looks quite different now. Earlier, I've seen the same errors there. > The armhf & ppc64el buildd failures I can't reproduce on the > porterboxes, although the latter is likely another flaky test. armhf likely needs to be retried as the build timed out. > > However, in local builds for unstable/amd64 in sbuild, I didn't get a > > single successful build. > > If it's not any of the mentioned failures, then some more information > would be useful, as I have quite the opposite behavior locally. What information are you interested in? I also tried in pbuilder and there I only got the utf_char2bytes failure, not the ones with check_cores. This hints that something about my build environment is fishy. If I understand correctly, my user's resource limits "leak" through sbuild while they don't pass through pbuilder. Would limiting the number of processes or virtual memory be a plausible explanation for the failures? So it might be useful to reduce the scope of this bug to the utf_char2bytes failure and the armhf/ppc64el buildd failures. Helmut
