ghc: patch readthedocs layout to work with our sphinx

[amadaluzia]

Testing the changes

• I tested the changes in this PR: partially (I do not have any Void Linux machines to test it on, this was confirmed to build within an Arch Linux container)

Local build testing

• I built this PR locally for my native architecture, x86_64-glibc
• I built this PR locally for these architectures (if supported. mark crossbuilds):
  • x86_64-musl
  • i686

Thanks @JkktBkkt for giving me this patch. This is the first step to closing #59852, the next will be updating GHC.

[amadaluzia]

Realised just now I mislabled the commit. I'll wait for builds to finish to correct it though. I'm unsure if marking this as a draft will automatically cancel the CI.

[JkktBkkt]

I've tried building pandoc with this, and while it worked on x86_64 and i686, x86_64-musl fails to for me, getting stuck after building a single library. Any ideas?

basement                         > [81 of 84] Compiling Basement.Terminal.ANSI
basement                         > [82 of 84] Compiling Basement.From
basement                         > [83 of 84] Compiling Basement.Block.Builder
basement                         > [84 of 84] Compiling Basement.String.Builder
basement                         > copy/register
basement                         > Installing library in /builddir/pandoc-3.6/.stack-work/install/x86_64-linux/8de8ccbdb267e392fa678f8d405fd39fc57eb2ca1dbf405d4a28fed11663c449/9.8.4/lib/x86_64-linux-ghc-9.8.4/basement-0.0.16-CvYYRhEW2X76ON6u3UsrAC
basement                         > Registering library for basement-0.0.16..

Not sure why this happens, chose pandoc specifically because it's one of the packages needed for libffi update

Have you tried building something else? Is this an issue with me building both on the same machine, no worktree, etc.?

[amadaluzia]

That's an interesting issue to come across. Setting up a Void Linux container will be a bit annoying with systemd-nspawn, but I'll see what I can do with other containers, specifically podman.

[JkktBkkt]

Hm, I let it sit for a longer while on that and it just continued. Didn't look like there was any activity in top view for the processes spawned by that launch of xbps-src though.
docker/podman should work just fine, and for testing purposes you'd want to have it available. Ofc a VM is an option too.

edit: turns out it was networking related.

[amadaluzia]

Either way, it's probably worth installing a Void Linux VM on my end. Sadly, podman and docker containers are really frustrating to work with when you aren't messing with /^(Docker|Container)files$/ or docker-compose. I'll see if QEMU can do the trick.