| 11 Jan 2025 |
 alexfmpe | hmm, well I'm overriding on configuration-common.nix, but 9.8 is the default now on haskell-updates
are you saying there's a bootstrap package set separate from haskellPackages ? | 04:44:58 |
 Alex | - Fetch bindist
- Build Hadrian using bindist
- Build GHC using bindist
- Build selected packages
Steps 2 and 4 will produce different derivations for each package. You should find that they have different GHC versions in their inputs. | 04:48:21 |
| Alex | TL;DR: Yes. | 04:50:28 |
 sterni (he/him) | alexfmpe: there's no separate bootstrap set which is why it's rebuilt | 13:18:31 |
| sterni (he/him) | I've tested a fix for happy already, will push that in a second | 13:30:25 |
| sterni (he/him) | https://hydra.nixos.org/eval/1810967 | 14:30:33 |
|  SigmaSquadron changed their display name from SigmaSquadron (Away until 2025-01-11) to SigmaSquadron. | 15:06:54 |
| alexfmpe | nice, going to get miso/reflex-dom building again - there's a few patches now failing from upstreaming and whatnot | 15:19:35 |
| alexfmpe | psqueues is a dep of both and also hitting the tasty-quickcheck issue, added a jailbreak next to your hashable one | 15:20:06 |
| alexfmpe | rebased your recent pushes, rebuilding hackage again | 15:20:23 |
|  oak 🏳️🌈♥️ changed their profile picture. | 16:44:58 |
| oak 🏳️🌈♥️ removed their profile picture. | 16:46:07 |
| oak 🏳️🌈♥️ set a profile picture. | 16:46:35 |
| alexfmpe | for some reason, this #if evaluates to true when building haskellPackages.reflex from haskell-updates, but not when building with cabal
https://github.com/reflex-frp/reflex/blob/b1767c47b47aae2e314166b1c6f4adefeb0e5128/src/Data/AppendMap.hs#L33
I'm using witherable 0.5 in both cases | 18:32:56 |
| sterni (he/him) | cabal-install is known for circumventing the prepolutade package db are you sure it's the same version in both cases? | 19:12:35 |
| alexfmpe | hmm I saw witherable-0.5 getting downloaded/built by cabal, and haskellPackages.witherable comes out as 0.5 as well | 19:52:34 |
| alexfmpe | notably, it's the nixpkgs version that's doing the unexpected thing and including those imports | 19:53:13 |
| alexfmpe | how can it fail the (0,3,2) check? | 19:53:33 |
| alexfmpe | it's acting as if witherable < 0.3.2 | 19:53:59 |
| alexfmpe | even weirder, the .cabal has witherable > 0.4 in there | 19:54:53 |
| alexfmpe | so this should be dead code in the first place | 19:55:03 |
| alexfmpe | related question: I submitted a PR removing it, but over on nixpkgs keep failing to apply the other hackage->master patches so this one doesn't get conflicts
do we overrideSrc or wait for a release? | 19:56:38 |
| alexfmpe | * even weirder, the .cabal has witherable >= 0.4 in there | 20:04:00 |
| alexfmpe | actually nevermind, managed to workaround with a simple postPatch sed, good enough until the release | 20:33:02 |
| sterni (he/him) | alexfmpe: ig you could generate a cabal freeze file to verify?! | 21:46:00 |
| alexfmpe | on the non-nixpkgs build? | 21:46:31 |
| sterni (he/him) | yes | 21:48:19 |
| 12 Jan 2025 |
|  Rayane Nakib (ريّان نقيب) joined the room. | 12:40:13 |
| 13 Jan 2025 |
 m1-s | Hi, i am working on a project where I just updated from nix 22.11 to 24.11. this project uses a haskell library that does ffi stuff. since the update I am getting the following error when trying to compile something that uses the ffi haskell library:
> <no location info>: error:
> /nix/store/ig69mqspa9a7gmkib1p05rndsfmp5qjl-mypackage-0.1.0/lib/ghc-9.6.6/lib/x86_64-linux-ghc-9.6.6/mypackage-0.1.0-GG7CxsqkGDRPE7u83aSVw-ghc9.6.6.so: undefined symbol: my_function
Anyone has a clue what broke or how to debug this? Did anything change in the way nix/ghc handles ffi things since nix 22.11?
| 08:05:19 |
| m1-s | it builds in the dev shell with cabal. just not in nix. so i must be missing some flags in my nix build. dont know which though. | 08:17:12 |
