| 2 Apr 2025 |
hellwolf | Thank you for all your work :) | 17:08:06 |
| tek changed their display name from torsten to tek. | 18:06:46 |
| 3 Apr 2025 |
hellwolf | In reply to @alexfmpe:matrix.org https://github.com/NixOS/nixpkgs/pull/395339 so, with that branch lsp is working for ghc 912 now? I will give it a try. That stopped me from updating my project from switching to 9.12.
Btw, would 9.12.2 make it there too? | 06:24:53 |
hellwolf | * so, with that branch, HLS works for ghc 912 too? I will give it a try. That stopped me from updating my project from switching to 9.12.
Btw, would 9.12.2 make it there too?
| 06:25:27 |
Alyssa Ross | In reply to @sternenseemann:systemli.org Alyssa Ross: GHC supports being bootstrapped with the two major versions before it officially (so 9.8 can bootstrap 9.10 and 9.12). Self bootstrapping often works though it probably requires some tinkering. It's probably worth investigating whether 9.8 can bootstrap 9.8 since we're currently upgrading to that version. Looks like it just works! I'll open a PR and we can figure out details :) | 10:58:57 |
Alex | NB: Isn't self-booting supposed to work for every version?
I don't see how stage 2 could be compiled if stage 1 can't be compiled by an identical stage 0. | 11:30:11 |
alexfmpe | In reply to @hellwolf:matrix.org so, with that branch, HLS works for ghc 912 too? I will give it a try. That stopped me from updating my project from switching to 9.12. Btw, would 9.12.2 make it there too? That PR has been merged, so you want haskell-updates branch really. Can't say I've tried HLS 2.10 yet and there were also a couple reports of seg faults over at #haskell-language-server:matrix.org so there might even be a 2.11 soon | 12:06:54 |
alexfmpe | * That PR has been merged, so you want haskell-updates branch really. Can't say I've tried HLS 2.10 yet and there were also a couple reports of seg faults over at #haskell-language-server:matrix.org so there might even be a new release soon | 12:47:38 |
alexfmpe | also, haskell-updates has 9.8 as default ghc, but 9.12.2 is there and the default ghc912 | 12:49:04 |
alexfmpe | I'm not sure exactly how/when the caching jobs work, but I expect as long as 9.12 isn't the default, you'll have cached compiler but not cached packages | 12:50:11 |
Alex | In reply to @alexfmpe:matrix.org I'm not sure exactly how/when the caching jobs work, but I expect as long as 9.12 isn't the default, you'll have cached compiler but not cached packages Cache is populated from haskell-updates (and IIRC only from that branch). | 12:51:25 |
maralorn | Well and from master of course. Although the list of jobs which get build on which branch is slightly different. | 13:03:15 |
maralorn | Some central packages, e.g. hls are being cached for every (major) compiler version. | 13:03:46 |
alexfmpe | noice | 14:00:41 |
alexfmpe | yeah this is what confuses me, which does which | 14:00:55 |
alexfmpe | I haven't gotten even 9.8 from cache on my mac on haskell-updates | 14:01:13 |
alexfmpe | * I haven't gotten even 9.8 from cache on my m1 mac on haskell-updates | 14:01:39 |
alexfmpe | * I haven't gotten even ghc 9.8 from cache on my m1 mac on haskell-updates | 14:02:11 |
maralorn | Well builds also need to succeed of course. | 14:21:22 |
maralorn | well compare pkgs/top-level/release.nix (ctrl+f haskell) with pkgs/top-level/release-haskell.nix | 14:26:42 |
Profpatsch | Is there a shortcut for building an executable with pre-set GHC RTS params? | 14:26:49 |
Profpatsch | In particular -Iw60 for small scripts to reduce the idle gc frequency | 14:27:04 |
maralorn | Yes, you can pass RTS flags to set by default to ghc. | 14:27:27 |
maralorn | e.g.: -with-rtsopts=-maxN2 | 14:28:01 |
maralorn | * e.g.: ghc-options: -threaded -with-rtsopts=-maxN2 | 14:28:28 |
Profpatsch | Ah you mean in the cabal file? | 14:28:43 |
Profpatsch | I guess that works | 14:28:46 |
maralorn | According to that we mainly build every ghc major release+hls on master and plenty more on haskell-updates. | 14:30:16 |
alexfmpe | Huh so the darwin build is probably failing is that it? | 14:42:16 |
maralorn | Oh, no. | 14:45:15 |