Hello, zimoun <zimon.touto...@gmail.com> skribis:
> On Wed, 13 Nov 2019 at 14:42, Ludovic Courtès <l...@gnu.org> wrote: > >> zimoun <zimon.touto...@gmail.com> skribis: >> >> > From my test with Guix d258d9c7d222e6b64531c14293f41bd8d62ea4f7, >> > "guix challenge guitarix" and "guix build --rounds=3" do not report >> > issues about reproducibility. >> > >> > And from my knowledge, the waf-based packages are not affected. >> > >> > Do you agree to close this bug since it is not relevant anymore? >> >> If you’ve checked that a local build gives the same result several times >> in a row (make sure it actually rebuilt things; “guix build --rounds=3 >> foo” does nothing if “foo” is already in the store), then you can >> definitely close it! > > I think I did but how do you do that cleanly? > > > > I am testing with "brute force" method: "guix gc -C" then proceed. I > notice unexpected behaviour; not sure it is the right place to report. I would probably do “guix build guitarix” (get the substitute) and then “guix build guitarix --check --no-grafts -K”, possibly several times*, which builds guitarix alone from source. (*) ‘--rounds’ is ignored when combined with ‘--check’, go figure… > 1. The first issue is that the two following commands do not populate > the same way. > > (Because my machine is not very powerful, before building I populate > the store with the dependencies from substitutes.) > > $ guix environment guitarix -- echo Done That should work. > $ guix build `guix show guitarix | recsel -R dependencies` That gives a different result because ‘dependencies’ does not show implicit inputs. > Why `ghostscript` is downloaded and not reported; neither by the two > populating commands? This is because of grafts, which lead to a poor UX as things are… Perhaps there are genuine issues described in the rest of your message but we’d need to isolate them first. :-) Thanks, Ludo’.
