| 1 Jun 2021 |
 maralorn | * sterni (he/him): You have two weeks. Don‘t let anything put you under pressure. | 07:49:03 |
|  bflyblue joined the room. | 08:23:53 |
 Las | How would I set GHCRTS when building for all haskell packages? | 09:40:49 |
| Las | Trying to reduce RAM usage | 09:41:08 |
| maralorn | Las: I am honestly not sure what the best approach is. But one approach I have seen is doing an overlay for haskellPackages, where you override mkDerivation to pass that argument to GHC. Look at the generic-builder.nix in the haskell-modules folder to understand what the build does and what args you need to pass. Don‘t forget that this will mean a lot of rebuilds. | 09:45:03 |
| Las | Thanks, yeah | 09:45:25 |
| maralorn | You probably only want to do this to the packages you don‘t get from cache. | 09:45:44 |
| Las | My problem is that aeson (IIRC) doesn't build on aarch64 with 4 GB of RAM | 09:45:54 |
| maralorn | But I know now trivial heuristic for that. | 09:45:57 |
| maralorn | But is aeson-aarch64 broken in nixpkgs? | 09:46:25 |
| Las | Haven't checked | 09:46:44 |
| maralorn | I mean if you can substitute it you don‘t need to build it, right? | 09:47:05 |
 sterni (he/him) | https://hydra.nixos.org/build/144126980 | 09:47:09 |
| Las | に返信 @maralorn:maralorn.de
I mean if you can substitute it you don‘t need to build it, right? I'm using GHC 9.0.1 | 09:48:41 |
| sterni (he/him) | oh | 09:48:50 |
| Las | So that's likely why | 09:48:48 |
| maralorn | Oh, okay. | 09:48:53 |
| Las | I can't find any job for it | 09:48:54 |
| maralorn | Yeah, we don‘t support that yet. Too much breakage … | 09:49:09 |
| Las | Wish the GHC team would make backward compatible releases like rustc | 09:49:38 |
| maralorn | Although, I am almost certain that e.g. hls depends on aeson so if we had that build hydra would at least try … | 09:50:50 |
| maralorn | But we recently disabled that job because hls is not supported on ghc 9 anyways. | 09:51:05 |
| maralorn | * But we recently disabled that job because hls is not yet supported on ghc 9 anyways. | 09:51:12 |
| sterni (he/him) | well the base update breaks more than the GHC update itself right | 09:51:22 |
| sterni (he/him) | working on fixing tests.writers for darwin currently | 11:34:32 |
| sterni (he/him) | seems like there are some severe darwin regressions for GHC, it seems to be incapable of producing any binaries unless it is in an environment with stdenv | 11:35:02 |
| sterni (he/him) | so it works within a nix-shell and a build which is probably why no one noticed | 11:35:15 |
| sterni (he/him) | we really should separate the GHC wrapper script and the build itself | 11:41:53 |
| sterni (he/him) | changing a line in the script requires a full GHC build | 11:42:04 |
| Las | I am following https://haskell4nix.readthedocs.io/nixpkgs-users-guide.html#how-to-build-a-stand-alone-project, and doing nix-shell as described doesn't give me a cabal. Is the guide outdated? I've updated my nixpkgs recently. | 13:31:40 |
