| 23 Apr 2025 |
 Teo (he/him) | You should always have a manual step where you check for existing PRs/issues that do the same thing. Also some maintainers don't want PRs and prefer issues for package bumps. It's tough to automate these things. In terms of automatic notifications, that's what stackage and hackage do already | 09:56:18 |
 hellwolf | I agree. It must be opt-in only. I receive often from GitHub PRs, but I can always opt-out | 10:32:40 |
 maralorn | alexfmpe: Huh. Merging the .jsexe PR into staging has the annoying effect that it will now take a while until it reaches haskell-updates … | 11:25:16 |
 alexfmpe | I mean we can cherry pick | 11:26:26 |
| maralorn | As you wish. I guess I am right now not affected by it. | 11:34:37 |
| alexfmpe | Me neither | 11:46:15 |
| alexfmpe | Getting it mainlined and cached removes the main obstacle | 11:47:03 |
| alexfmpe | By the time anyone is using it and wants to upstream some override we'd probably have it merged into haskell-updates | 11:47:41 |
 sterni (he/him) | Just cherry pick I have a feeling we may doing that a fair bit for 24.11 stuff | 11:54:52 |
| maralorn | 24.11? | 12:33:40 |
| 24 Apr 2025 |
|  @jktr:0x16.de left the room. | 01:16:55 |
|  Leon Vatthauer set a profile picture. | 12:50:31 |
|  Peter Dragos joined the room. | 14:13:49 |
| Peter Dragos | Hey folks, I'm working on getting haskellPackages.accelerate building. I've gotten it building locally against 24.11.
I'd like to make a PR for nixpkgs, so I've checked out nixpkgs/haskell-updates and am editing common-configuration.nix with
accelerate = overrideSrc {
src = pkgs.fetchFromGitHub {
owner = "AccelerateHS";
repo = "accelerate";
rev = "3f681a5091eddf5a3b97f4cd0de32adc830e1117";
sha256 = "sha256-tCcl7wAls+5cBSrqbxfEAJngbV43OJcLJdaC4qqkBxc=";
};
} super.accelerate;
But when I do this, I get
double-conversion >=2.0, formatting >=7.0, microlens >=0.4
if I go into the nix repl, :lf, and inspect nix-repl> legacyPackages.x86_64-linux.haskellPackages.microlens.version I get "0.4.13.1" which should be compatible; and this is the same for double-conversion and formatting as well. Do I need to do something different to get the dependencies picked up properly?
| 14:14:29 |
|  TuXic joined the room. | 14:21:59 |
| maralorn | Peter Dragos: Its likely that those dependencies were not present in the version of accelerate which is in that nixpkgs checkout. You are reusing the derivation for another version of the package there, just swapping the src. | 14:37:35 |
| maralorn | Is there no newer release of accelerate? | 14:38:10 |
| maralorn | A possible solution is to manually add the missing dependencies with further overrides on the package. | 14:39:43 |
| Peter Dragos | I far as I can tell, the latest version of accelerate on hackage has the same base constraints | 14:43:58 |
| Peter Dragos | I'll give this a shot, thanks | 14:44:06 |
| maralorn | Why are you mentioning base constraints? They don’t appear in your error message. | 14:44:58 |
| maralorn | Huh, I see your point. Last release 5 years ago. Unclear whether you can compile this against a current ghc. | 14:47:56 |
| Peter Dragos | Sorry, I originally asked over in the Haskell matrix and then realized it was better off here.
The relevant info is that the version in release-24.11 gives
> Error: Setup: Encountered missing or private dependencies:
> base >=4.12 && <4.15
>
| 14:48:28 |
| maralorn | Well then first you try doJailbreak. | 14:49:12 |
| maralorn | If it works it’s the easiest solution. But it probably doesn’t … | 14:49:39 |
| Peter Dragos | But the repository itself (link] is active. Jailbreaking didn't work, so someone suggested pulling directly from the repo. I got that to build on 24.11 with one other dontCheck on a dependency, but haskell-updates has the issue mentioned above | 14:50:07 |
| Peter Dragos | * But the repository itself (link) is active. Jailbreaking didn't work, so someone suggested pulling directly from the repo. I got that to build on 24.11 with one other dontCheck on a dependency, but haskell-updates has the issue mentioned above | 14:50:19 |
| maralorn | Huh, then I am confused. | 14:50:50 |
| maralorn | Because my explanation should also mean that it doesn’t work on 24.11. Since the derivation clearly hasn’t changed for years. | 14:51:20 |
| maralorn | Just to be sure: The error message appears in the build of accelerate and not in one of its dependencies? | 14:52:02 |
