| 17 Oct 2024 |
 Arian | Yes | 07:25:47 |
| Arian | Like a factor 20-100 on my network | 07:25:59 |
 emily | not that bad for me, but still significant | 07:26:44 |
 K900 | Github is faster for me sometimes but my internet is giga fucked | 07:27:17 |
| Arian | We could in theory point the registry to channel uris right? | 07:27:25 |
| K900 | So my opinion shouldn't matter | 07:27:25 |
| emily | In reply to @arianvp:matrix.org
We could in theory point the registry to channel uris right? we'd lose the Git .rev type information I think | 07:28:23 |
| emily | though Nixpkgs can fall back to the information Hydra sneaks into the tarball | 07:28:36 |
| Arian | https://releases.nixos.org/nixpkgs/nixpkgs-24.11pre693341.7881fbfd2e3e there's a git-revision file | 07:28:59 |
| Arian | Just need to teach nix about it | 07:29:08 |
| emily | right. Nixpkgs knows about it but not Nix, AIUI. | 07:29:51 |
| emily | which means potentially turning a simple PR to the registry into Nix feature development 🫠 | 07:30:09 |
| emily | (but perhaps it doesn't matter) | 07:30:29 |
 vcunat | In reply to @k900:0upti.me
Github is faster for me sometimes but my internet is giga fucked If Fastly becomes the default for this, it will probably get faster than now thanks to caching. | 08:02:40 |
 SomeoneSerge (utc+3) | In reply to @emilazy:matrix.org
which means potentially turning a simple PR to the registry into Nix feature development 🫠 H'm I'm generating a static registry.json from my repo's pins on all my devices and I honestly don't see why would "going online" ever make sense as a default. We could just make that into a nixos module, and then it's not nix feature development but a nixpkgs PR | 08:32:28 |
| SomeoneSerge (utc+3) | Not sure how to go around the side-effect of dropping all the non-nixpkgs stuff from the json | 08:34:08 |
| emily | NixOS already pins nixpkgs flake out of the box these days | 08:34:13 |
| emily | (in a way that's kind of bad and broken) | 08:34:19 |
| emily | fwiw if you do that it makes locking nixpkgs inputs without an explicit GitHub reference act weirdly (e.g. I committed a Nix store path to the nix-darwin flake.lock for several months) | 08:34:56 |
| emily | In reply to @emilazy:matrix.org
(in a way that's kind of bad and broken) (though to be clear I don't know how much better you can really do) | 08:35:11 |
| SomeoneSerge (utc+3) | In reply to @hexa:lossy.network
sent an image. I paused to think about how does one even a/b rate limits | 08:38:15 |
| SomeoneSerge (utc+3) | By the way is it by design that deleting a discourse account necessarily entails deleting the messages? | 08:46:55 |
 hexa (signing key rotation when) | it entails anonymizing the messages | 09:33:03 |
| Arian | This is about the revision info in the lock file? | 09:54:45 |
| Arian | Idk if it's that important. And in the tarball we override the nix version functions to return the revision iirc | 09:55:02 |
| Arian | Due to there being a .nix-version file in the src | 09:55:17 |
| Arian | So only the lock file will suffer. But any nixos and nix code still works | 09:55:38 |
 Gaétan Lepage | Redacted or Malformed Event | 10:59:52 |
| hexa (signing key rotation when) | Gaétan Lepage: pretty sure nvcc and pt_main_thread processes are at fault | 11:08:38 |
| hexa (signing key rotation when) |
Out of memory: Killed process 817379 (python3.12)
| 11:08:52 |
