Edouard Klein <[email protected]> writes: >> There’s been frequent breakage over the past couple of months. That’s >> why I think having continuous integration for pull requests is top >> priority for the project. >> > > That's above my current level of spare capacity both in hardware and in > development time, however, do you think that, in the meantime, there > would be value in knowing the latest commit for wich all packages build > ? Is this something that's already known with the current infrastructure ?
We never had zero build failures, but that would be a goal. To me the first milestone would be to prevent the introduction of new build failures, which was the motivation for https://pulls.ci.guix.gnu.org: https://lists.gnu.org/archive/html/guix-devel/2025-09/msg00153.html >> A distro that breaks this often is barely viable; if even we, people who >> follow it pretty closely, have a hard time coping with it, imagine what >> it’s like to someone who’s not contributing to Guix and/or following it >> from a distance. >> > > While true, I think we also should be careful not to slow the develpment > pace too much. I think we can have our cake and eat it too. :-) > But I do agree that a build check on PRs would be nice. But it will be > very costly to run e.g. a perl or python update and we'd need to rebuild > half the world. Yeah, these wide-reaching changes are of course harder to deal with than day-to-day changes because they can lead to breakage at a distance. And yet, as you write, we want to be able to have such changes rolling at a reasonable pace. Thanks, Ludo’.
