In reply to @maralorn:maralorn.de
Tristan Ross: I don’t think that has anything to do with the LLVM bump stuff, which seems relatively independent anyway. I can’t judge that PR, but the problem with it is figuring out whether it works. If it does work we can basically merge it at any time, independently of the following:
I want to highlight a different possibility: sterni and I settled on the deprecation policy, which I will announce (with a request for feedback) shortly. With that settled we are free to remove multiple old ghc versions once all in-tree consumers have been resolved. I guess the only point of contention is that we didn’t announce those deprecations in the last release notes after all, so if we want to stay true to that policy we probably should aim to drop them directly after 25.05 and announce the deprecation in 25.05. I know this is slower than you wanted this. But at least it is a solution. Would that be okay?

the problem with it is figuring out whether it works. If it does work we can basically merge it at any time

If I spin up a build of it on my ARM hardware and it passes, is that good for it to merge? The PR simply moves LLVM 12 to 13 and 15.

I want to highlight a different possibility: sterni and I settled on the deprecation policy, which I will announce (with a request for feedback) shortly.

That'd be great

we probably should aim to drop them directly after 25.05 and announce the deprecation in 25.05. I know this is slower than you wanted this. But at least it is a solution. Would that be okay?

If we can at least try and get my PR in, I think it'll be a good compromise. 24.11 has fully working GHC, 25.05 would have the few on 12, 25.11 would drop the old ones from the sound of it.

       … while evaluating the attribute 'nixops_unstablePlugins.nixops-aws'
         at /nix/store/zkpypzhg52sqw7jc70l7hgfys002rbgc-source/pkgs/applications/networking/cluster/nixops/default.nix:38:9:
           37|
           38|         nixops-aws = throw "nixops-aws was broken and was removed from nixpkgs";