| 25 Nov 2024 |
 alexfmpe | I was going to see something in ghcup, but a dep didn't build out of the box with nix-shell -p
wondering where to branch from when attempting a fix | 23:03:25 |
 sterni (he/him) | if the rebuilds are not expensive and it's not going to cause nightmare merge conflicts with haskell-updates you can merg into master | 23:25:01 |
| alexfmpe | I feel like we'll almost always have conflicts if overrides tend to get added at the end of configuration*.nix files | 23:56:15 |
| 26 Nov 2024 |
| sterni (he/him) | well that's the same with working on the branch—we always try to tell people to add overrides in a random place | 13:42:30 |
| alexfmpe | Working on the branch means the PR author has to fix the conflict | 15:51:08 |
| alexfmpe | Otherwise the PR maintainer does on their merge | 15:51:25 |
| alexfmpe | Adding overrides in random places for the sake of less conflicts feels weird to me.
Wouldn't sorting them lexicographically solve that? | 15:52:54 |
| sterni (he/him) | you can also do that of course yeah, but the files aren't sorted lexicographically at the moment | 16:21:19 |
| sterni (he/him) | and there are good reasons to sometimes group things thematically which is done in an ad hoc fashion | 16:21:34 |
 emily | we also place release notes randomly (in theory) | 16:23:37 |
| emily | (in practice…) | 16:23:39 |
| emily | the real solution is to split things into multiple files tbh | 16:23:51 |
| emily | Nix is on one release note per file and NixOS will hopefully be at some point | 16:24:00 |
| emily | it's marginally more Git-storage-friendly too (though packfiles complicate that) | 16:24:18 |
| sterni (he/him) | well then you have the problem with directories because the tree objects become huge at some point | 16:26:56 |
| sterni (he/him) | we kind of missed that with by-name where we only thought about the github rendering limit of 1k entries | 16:27:27 |
| sterni (he/him) | but apparently git tree objects already become kind of bad at >300 entries | 16:27:40 |
| sterni (he/him) | * but apparently git tree objects already become kind of bad at ~300 entries | 16:27:47 |
| emily | time for three-character prefixes? | 17:26:21 |
| emily | I guess that's bad because the by-name directory would become huge :) | 17:26:42 |
| emily | so I suppose we need pkgs/by-name/ab/c/abcd or something | 17:27:00 |
 iqubic (she/her) | Hmm... I'm trying to compile a Haskell project with Nix and I'm getting an error related to Finite-Typelits. | 17:41:23 |
| iqubic (she/her) | This is the output from nix log: https://dpaste.com/E5LC6FTTC | 17:44:34 |
| iqubic (she/her) | I'm just telling Nix to use Cabal2Nix with this cabal file: https://dpaste.com/FY4FTT7VB | 17:48:11 |
| iqubic (she/her) | Finite-Typelites should be using version 0.2.1.0 | 17:48:48 |
| iqubic (she/her) | * Finite-Typelites should be using version 0.2.1.0, but it isn't. | 17:48:54 |
| iqubic (she/her) | Actually, according to https://hackage.haskell.org/package/finite-typelits NixOS is still using 0.1.6.0 | 17:49:57 |
| iqubic (she/her) | Annoying... | 17:50:01 |
 maralorn | In reply to @emilazy:matrix.org
so I suppose we need pkgs/by-name/ab/c/abcd or something Multi charakter filenames where a mistake. "pkgs/by-name/l/i/b/c/libc" | 17:54:25 |
| maralorn | In reply to @emilazy:matrix.org
so I suppose we need pkgs/by-name/ab/c/abcd or something * Multi-character filenames where a mistake. "pkgs/by-name/l/i/b/c/libc" | 17:54:33 |
