| 8 Aug 2025 |
 chreekat | It's an excellent post | 11:59:28 |
|  @selkie-diligent-61ad45186da03739848be880:gitter.im left the room. | 13:55:04 |
 emily | has anyone noticed a GHC build hanging around building ghc-9.8.4 (buildPhase): | Run Ghc CompileCWithGhc Stage1: rts/Hash.c => _build/stage1/rts/build/c/Hash.thr_debug_dyn_o with a bunch of C compilers using all the CPU they can find overnight but seemingly not progressing | 14:11:44 |
| emily | (I suspect the answer is no and this is an LLVM 21 regression but I thought I'd check) | 14:11:53 |
 maralorn | sterni: Excellent post. I found two typos in the paragraph
When you’ve found a fix for a package, make sure to add a comment next to your override explaining explaining why it is needed and ideally link an issue or PR that needs to be resolved upstream for us to drop the override (if no issue exist, do open one).
Their is a double "explaining" and "no issue exist" is missing numerus congruency.
| 16:38:06 |
| maralorn | * sterni: Excellent post. I found two typos in the paragraph
When you’ve found a fix for a package, make sure to add a comment next to your override explaining explaining why it is needed and ideally link an issue or PR that needs to be resolved upstream for us to drop the override (if no issue exist, do open one).
There is a double "explaining" and "no issue exist" is missing numerus congruency.
| 16:38:39 |
 sterni | emily: nope and I have compiled a quite a few GHCs on haskell-updates the last week | 17:07:09 |
| sterni | though x86_64-linux so no LLVM involvement at all | 17:07:37 |
| 9 Aug 2025 |
|  crtschin joined the room. | 01:58:10 |
|  danl joined the room. | 05:01:30 |
| emily | how does one understand why a non-Stackage package is several major versions old in the Haskell package set? would it be because of things requiring an older version, or are Hackage packages not bumped en masse at all? | 22:03:17 |
| emily | llvm-tf is at 16.0 and I don't immediately see why – seems like the stuff pinning llvm-* to old versions are already marked broken | 22:04:57 |
| sterni | because the llvm-tf > 16 versions have only been uploaded a month ago | 22:08:41 |
| sterni | staging is on 21.0 | 22:08:47 |
| emily | ah I didn't notice that even 17.0 was so recent… thank you | 22:10:26 |
| sterni | you should be able to find out by grepping pkgs/development/haskell-modules/configuration-* if it should be the case | 22:11:30 |
| emily | is it intentional that https://github.com/NixOS/nixpkgs/blob/ca321b05ae7f096c195eb14809e55ef9921f19eb/pkgs/development/haskell-modules/configuration-nix.nix#L506-L514 will make CI fail on LLVM bumps? | 22:11:40 |
| sterni | yes-ish | 22:12:14 |
| sterni | it's mostly meant to fail our CI, but it's hard to implement that | 22:12:42 |
| emily | right | 22:14:00 |
| sterni | in my experience bindings etc. are pretty tricky because no one tests these when updating the base package, so I wanted to make it visible for us here at least | 22:14:35 |
| sterni | but we should probably revert it | 22:14:49 |
| sterni | the problem is also that you can't just update the number on the bump yet because the package wants to link against libLLVM21-git | 22:15:33 |
| sterni | at least last I checked | 22:15:40 |
| emily | well, we do build quite a lot of stuff when bumping LLVM | 22:32:12 |
| emily | at least | 22:32:14 |
| emily | so it may be a special case in terms of bindings | 22:32:17 |
| 10 Aug 2025 |
| sterni | Haskell bindings get frequently overlooked in my experience | 00:44:17 |
| sterni | though we haven't had Haskell LLVM bindings where an up to date version is readily available | 00:44:43 |
| sterni | llvm-hs never uploaded anything after 9.0.1 to Hackage and I never bothered to work out how usable the llvm-12 branch was | 00:45:40 |
