lambdabot: apply upstream patches for GHC 9.0.2 support

[ncfavier]

Tested by building lambdabot and running dice d10 (from the dice plugin) and yow (from the fortune plugin) at the prompt.

[maralorn]

Thank you for your fixes, especially for creating the upstream PR!

1. normally we apply overrides for the current default compiler in configuration-common.nix
2. we prefer applying patches instead of switching to unstable, if possible. Especially in this case where the used commits are not even on upstream master.

Can you change the PR to accommodate that, or explain why that’s not feasible?

[ncfavier]

Done (and tested again). Wasn't sure where to put my changes in configuration-common.nix so I picked a random place.

[maralorn]

Okay that’s already better.

Now, there is one thing I should have warned you about. We try to not accumulated patches in nixpkgs, but rather use pkgs.fetchpatch to pull the patch from github. There are various examples in configuration-common.nix.

Since you are patching different packages in the same commit, if I am not mistaken, this would be a bit finnicky with fetchpatch.
You would need a combination of the stripLen = 2, extraPrefix = "" and includes = [ ... ] or something like that. And be careful since this is a fixed output derivation you will need to invalidate the hash everytime you modify one of those parameters.

Do you want to try that? If it’s too annoying, I think I could be convinced to merge this as is …

[ncfavier]

I managed for dice.

For misfortune the PR is based on master which has a few more commits than the latest hackage release, and the patch doesn't apply cleanly. We would need to fetch two more commits from master for the patch to apply. I also can't just excludes = [ "misfortune.cabal" ] because the patch adds a dependency to the cabal file.

For lambdabot, what you suggest doesn't seem possible because the includes logic happens after the stripLen logic for some reason. Unless we make our own fetchpatch with fetchurl + runCommand.

Note that in both cases (misfortune and lambdabot), the only commits on top of the latest release are adjustments to CI, documentation or cabal files, so maybe we could fetch the source from the PR as I did initially and keep the version as-is since there are no code changes.

[maralorn]

Yeah, the stripLen/excludes order is really annoying.

Thank you for trying. I guess we should choose our battles wisely and not sink any more work into a working fix.

[ncfavier]

I guess we should choose our battles wisely and not sink any more work into a working fix.

I am not a very wise person #165327.

(Let's not block this PR on that one though)

[maralorn]

thx for patching