Hi, Maxim Cournoyer <maxim.courno...@gmail.com> skribis:
>> Thanks for the patch, that should break the deadlock and allow us to >> proceed with the release! >> >> Next we need to update the ‘release’ target so >> GUIX_ALLOW_ME_TO_USE_PRIVATE_COMMIT is set. > > Done! Thanks! >> the extra dependency on Git, > > To me, this script (update-guix-package), is an extension of the > Make-based build system (that's currently it's sole entry-point). There > are already calls to git in this build system (for example, to get the > commit corresponding to HEAD), so I don't perceive it as nasty in this > context. It can also be used as a reminder of things that are missing > in (guile git), for the purists ;-). I think we had everything needed in Guile-Git. >> the extra copies of the whole tree, > > There used to be 3 copies required in total (the Guix checkout, a first > dummy copy in the store, and a final copy in the store). > > Now, we have 2 copies unless GUIX_ALLOW_ME_TO_USE_PRIVATE_COMMIT is set, > in which case we get a third one in the store. Seems pretty even to me! Right, I stand corrected. >> the extra code > > The extra code make things smoother (better/faster diagnostic), causes > less friction in the workflow (I don't need to go paranoid about my tree > being in pristine condition before 'make update-guix-package' -- I can > rely on Guix computing it deterministically from the last commit). > >> and the shell pipelines, something avoided in the rest of Guix. > > Again, to me this script is a standalone extension of the build system. > It's not defined as a module, cannot be used in the rest of the code > base, so that it does things a bit differently doesn't strike me as bad. Note that quite a few modules started their life under build-aux/. >> Perhaps that suggests there are unwritten coding guidelines we should >> eventually discuss and write. We’ll see! > > That could be nice. Although a linter included with Guile (ala Rust or > Go) and configurable through a config file could have even more impact, > in my opinion. In any case I'd be honored that my code got to be the > spark behind such guidelines/tool, eh :-). I don’t think a linter could flag high-level patterns like the ones we’re talking about, but human-written text could. I hope this discussion can at least help improve mutual understanding on future patches and review processes. Thanks again for your time and patience! Ludo’.
