IRC channel logs

<rekado>about to push 2000+ commits to r-team :-/

<rekado>this is getting out of hand

<civodul>should it be a single commit?

<civodul>since we’re not testing intermediate stages anyway

<civodul>actually in what way is it getting out of hand?

<rekado>I actually tested hundreds of them :-/

<rekado>I've been at this for the past week.

<rekado>lots of packages needed to get a new arguments field to delete bad tests or disable vignette building

<rekado>I thought about doing it all in one commit, but the changes were a bit more involved and less generalizable, so I decided against it.

<civodul>oh ok

<civodul>ah yes, the change wrt. tests must have made that trickier than usual, right?

<civodul>i think we should look into full automation, similar to guix-cran

<rekado>yes, I think that would be good

<rekado>most CRAN packages are simple. We would need to have a way to define quirks to explicitly override whatever the automation gets wrong.

<civodul>yes

<civodul>someone would run “./pre-inst-env guix refresh -t cran -u && git commit -a”

<civodul>with the understanding that everything has already been built

<civodul>or something along these lines

<rekado>I haven't looked closely but I saw that an automated emacs channel that was mentioned on #guix last week had a quirks mechanism to fix up packages.

<civodul>doesn’t guix-cran also have something like that?

<civodul>though it’s quite different because everything’s regenerated every time

<rekado>guix-cran just uses the importer. It prunes some packages, but it does not attempt to support manual overrides.

<civodul>ok

IRC channel logs

2024-11-22.log