| 14 Oct 2021 |
 K900 | Same one here | 12:30:57 |
| K900 | ...that is weird | 12:31:02 |
| K900 | Is the CDN being weird again? | 12:31:11 |
 lukegb (he/him) | You sure you don't have a magic extra nixpkgs input that's tracking master somehow? :P | 12:32:10 |
| K900 | I definitely don't | 12:32:21 |
| lukegb (he/him) | What does e.g. curl -v https://cache.nixos.org/2zbj4ag44472w1g94k2fvzgp9i7jwwb9.narinfo get you? | 12:34:07 |
| lukegb (he/him) | Other dumb things: you haven't accidentally removed the cache.nixos.org binary cache? | 12:34:28 |
| K900 | Nope | 12:35:33 |
| K900 | It did fetch some things too | 12:35:39 |
| K900 | Just not all of them for some reason | 12:35:44 |
 moritz.hedtke | In reply to @lukegb:zxcvbnm.ninja
At that point you have a human doing the Hydra scheduler's job, and the time would be better spent fixing the scheduler, imo I thought hydra would use the nix distributed builders feature and do something like nix-build something/release.nix -A something --dont-abort-on-failed-builds or would that not work? | 12:41:57 |
| lukegb (he/him) | Alas, it's not quite that simple | 12:48:44 |
 Linux Hackerman | K900: the "advance-blocking set" for nixpkgs-unstable is pretty small, maybe try nixos-unstable instead? | 12:49:16 |
| lukegb (he/him) | Linux Hackerman: nixos-unstable won't block on Darwin at all, so that's probably not good advice for someone on Darwin | 12:49:45 |
| Linux Hackerman | Oh right never mind me, missed that bit of context, sorry | 12:50:03 |
| K900 | Yeah, I'm using nixos-unstable for my own boxes | 12:50:31 |
| Linux Hackerman | In reply to @k900:0upti.me
Just not all of them for some reason Maybe your "cache cache" is out of date? Try `--narinfo-negative-cache-ttl 1` | 12:50:48 |
| lukegb (he/him) | I half-wish we split the channels up by architecture... | 12:50:53 |
| Linux Hackerman | (Not 100% sure if I got the option name right, check nix.conf if it complains) | 12:51:12 |
| Linux Hackerman | man nix.conf* | 12:51:33 |
| Linux Hackerman | And if I'm missing context again I apologise and will shut up and go away. Or at least read some more scroll back. | 12:52:16 |
| K900 | At this point it's probably easier to wait for it to finish building tbh | 12:52:49 |
| lukegb (he/him) | Haha | 12:52:52 |
| K900 | But I'll try it next time | 12:52:53 |
| Linux Hackerman | Nix's behaviour of caching what the binary caches don't have can be a bit of a pain, though I can see why it does so | 12:53:52 |
 Domen Kožar | it's mostly done because of bad defaults :) | 15:50:10 |
| Domen Kožar | see https://github.com/NixOS/nix/pull/5324 | 15:50:35 |
| Domen Kožar | this speeds up narinfo query 4x if you have an 8 core machine | 15:52:25 |
| Domen Kožar | probably a magnitude if it's a 4 core | 15:52:33 |
| Domen Kožar | once you have this fixed, there's little reason to save a few ms of machine time at the expense of cache invalidation issues | 15:53:18 |
