| 7 Apr 2026 |
goldstein | reducing github request load 25x really did the difference | 16:50:02 |
goldstein | it also potentially allows benchmarking, although I didn’t get around to this yet | 16:50:29 |
accelbread | is self in flake registry that weird :P Isnt it common to set flake inputs to the registry in nixos config? That ends up setting self (which I use for nix develop self#rust or w/e | 17:06:17 |
accelbread | * is self in flake registry that weird :P
Isnt it common to set flake inputs to the registry in nixos config? That ends up setting self (which I use for nix develop self#rust or w/e) | 17:06:35 |
goldstein | you usually set, like, inputs.nixpkgs there, not inputs.self, no? | 17:07:00 |
zoë (she/her) | i heard about the great flake check a while ago and was really looking forward to the results, this is so cool :D | 17:07:34 |
accelbread | registry = mapAttrs (_: v: { flake = v; }) inputs; | 17:08:26 |
goldstein | yeah, i think it was an idiom like this actually | 17:08:54 |
goldstein | it’s my first time seeing this, I basically only use local registry for nixpkgs and my template set | 17:09:16 |
zoë (she/her) | this might be a dumb question, since i don't use flakes much, but: what's the difference between the registry and nix-channel? | 17:10:40 |
accelbread | Repos using my flake lib read their flake.lock to get inputs when not explicity passed, so interesting that there werent more failures due to that | 17:12:44 |
accelbread | In reply to @blokyk:matrix.org this might be a dumb question, since i don't use flakes much, but: what's the difference between the registry and nix-channel? afaik, channels are for syntax, and registries are for flakerefs | 17:13:30 |
goldstein | registry is a json file on your system (local registry) and another json file global for everyone (global registry) that is used to resolve flake references like nixpkgs | 17:13:44 |
goldstein | i.e. without an explicit url | 17:13:49 |
accelbread | * afaik, channels are for `<nixpkgs>`
syntax, and registries are for flakerefs | 17:13:54 |
goldstein | <nixpkgs> is resolved via NIX_PATH, not channels | 17:14:07 |
accelbread | oh, huh | 17:14:22 |
goldstein | channels is like a pre-flake way to distribute nixpkgs | 17:14:30 |
goldstein | I haven’t used them much, I migrated to flakes quickly, and then to unflake lol | 17:14:44 |
zoë (she/her) | yes sorry i knew that, what i meant is: aren't they both stateful/implicit ways of having dependencies? | 17:14:55 |
goldstein | yeah they ae | 17:15:06 |
goldstein | * yeah they are | 17:15:10 |
goldstein | maybe they failed for another reason earlier, maybe these repos just weren’t in the sample | 17:15:36 |
goldstein | maybe all the inputs were explicitly passed, so this code path was never reached | 17:15:57 |
zoë (she/her) | ok so what was the point of having the registry as a separate thing? ;-; | 17:16:18 |
zoë (she/her) | oooh wait is the registry only used at locking time? and not runtime? | 17:16:31 |
goldstein | nix channels are just for nixpkgs, registries can hold all sorts of stuff | 17:16:38 |
goldstein | the current global registry holds an unreasonable amount of random repos | 17:16:55 |
goldstein | for a reason I cannot comprehend | 17:17:02 |
goldstein | but you can also use it for local aliases and stuff | 17:17:12 |