| 4 Apr 2024 |
|  fabaff changed their display name from Fabian Affolter to fabaff. | 08:40:40 |
 @adam:robins.wtf | pardon if this was addressed in the by-name RFC, but was it every considered to have a packages.nix that could allow by-name to support multiple packages in a single directory? i'm still thinking over multi package support | 15:25:02 |
 infinisil | adamcstephens 🐝: Considered yes, but not thought out. Rather deferred for the future to keep it as small as possible for now | 15:31:59 |
| 5 Apr 2024 |
 Philip Taron (UTC-8) | infinisil: what do you think about having a function.nix indicate that what ought to exist in the by-name directory is a function, not a derivation? | 16:00:02 |
| Philip Taron (UTC-8) | * infinisil: what do you think about having a file named function.nix indicate that what ought to exist in the by-name directory is a function, not a derivation? | 16:00:13 |
 @piegames:matrix.org | In reply to @philiptaron:matrix.org
infinisil: what do you think about having a file named function.nix indicate that what ought to exist in the by-name directory is a function, not a derivation? I think this was initially attempted by the RFC | 16:02:14 |
| Philip Taron (UTC-8) | In reply to @piegames:matrix.org
I think this was initially attempted by the RFC I see https://github.com/nixpkgs-architecture/simple-package-paths/pull/34 which is where the restriction to just derivations is introduced, but I don't see a lot of justification for why nix-build -A name always builds is there. | 16:11:46 |
|  @yaqueen:matrix.org joined the room. | 20:35:04 |
| 8 Apr 2024 |
| infinisil | Personal office hour now: https://meet.jit.si/nixpkgs-architecture | 18:00:06 |
| infinisil | In reply to @philiptaron:matrix.org
infinisil: what do you think about having a file named function.nix indicate that what ought to exist in the by-name directory is a function, not a derivation? I don't think this was discussed actually. Personally I think we should have a clearer separation of derivations vs e.g. derivation-builders | 18:01:49 |
| infinisil | Dominic Mills: See you around, thanks for joining! | 19:53:41 |
| 9 Apr 2024 |
 willbush | I would have come, but I was viewing the eclipse that happened in the US. 😎 infinisil could you check out https://github.com/NixOS/nixpkgs/pull/302218. I was wondering if we could limit how often the actions? | 01:40:47 |
|  SomeoneSerge (back on matrix) changed their display name from SomeoneSerge (migrating synapse) to SomeoneSerge (void). | 13:24:41 |
| Philip Taron (UTC-8) | In reply to @infinisil:matrix.org
I don't think this was discussed actually. Personally I think we should have a clearer separation of derivations vs e.g. derivation-builders What would that cleaner separation look like? | 19:04:32 |
| infinisil | Philip Taron (UTC-8): Something like inherit (import <nixpkgs/api> {}) packageSets builders lib nixos ci; | 19:13:12 |
| Philip Taron (UTC-8) | But it wouldn't extend to what's passed in to callPackage-style files, though. | 19:13:48 |
| Philip Taron (UTC-8) | So lib would be passed in, substitute would be passed in, cargo would be passed in, even though one's a "module", ones a function generating a derivation, and one is a derivation | 19:14:30 |
| infinisil | Ah I see, I think this should be done too ideally! | 19:15:07 |
| infinisil | callPackage ({ fetchers, packageSets, ... | 19:15:35 |
| willbush changed their profile picture. | 22:42:42 |
| 11 Apr 2024 |
|  Anthony Rsl set a profile picture. | 21:59:17 |
| Anthony Rsl removed their profile picture. | 22:12:52 |
| 13 Apr 2024 |
|  terru joined the room. | 23:16:56 |
| 14 Apr 2024 |
 Robert Hensing (roberth) | lib could be a package; lib // { outPath = ../../lib; type = "derivation"; }. done. ;) | 15:23:25 |
| Robert Hensing (roberth) | In reply to @infinisil:matrix.org
callPackage ({ fetchers, packageSets, ... The "cost" of separating out the functions and package sets from the packages in pkgs, is that it exposes the flaw in callPackage that there's a missing indirection between "overridables" (the package function parameters) and their values, which, frequently, aren't pkgs.<name> | 15:26:39 |
| Robert Hensing (roberth) | Turns out I didn't actually support arbitrary overridables yet in single-fixpoint, but that was fixable (pun not intended): https://github.com/NixOS/nixpkgs/pull/296769/commits/b3eddddd4c268866f090934cf4c76ddec909b844 | 15:57:03 |
| Robert Hensing (roberth) | Commit message has an example of how an arbitrary overridable can be added, without touching all-packages.nix (!) | 15:58:00 |
|  @windchimesofmagic:matrix.org left the room. | 17:37:31 |
| 15 Apr 2024 |
| infinisil | Personal office hour now: https://meet.jit.si/nixpkgs-architecture | 18:00:06 |
| infinisil | Cc Philip Taron (UTC-8) willbush | 18:00:22 |
