16 Oct 2024 |
emily | compat stubs just weren't tested on Rust packages on Linux it looks like | 19:35:59 |
emily | because of the evil stuff the old SDK pattern needed to do for Rust in the first place | 19:36:14 |
| justscreamy joined the room. | 20:02:10 |
Randy Eckenrode | In reply to @emilazy:matrix.org K900: here you go https://github.com/NixOS/nixpkgs/pull/349117 Thanks for fixing. I’m surprised it got last ofborg. | 23:10:16 |
emily | AIUI it only breaks evaluation of the insides of Rust packages | 23:10:47 |
emily | I believe ofborg only catches eval failures that will "break eval", i.e. stuff that breaks before inside the body of whatever is callPackage d | 23:11:07 |
emily | but I'm not sure, since it also evaluates some tests and stuff… | 23:11:12 |
Lily Foster | Ofborg will catch failure to eval any outPath of hydra jobset | 23:14:10 |
emily | 🤔 | 23:14:47 |
Lily Foster | not sure how that could have gotten past ofborg either tbh | 23:14:55 |
emily | I think the answer is that it didn't actually finishing evaluating the final version because we rushed before the freeze, and IIRC on previous revisions there was some other eval-breaking thing. | 23:15:36 |
Lily Foster | * Ofborg will catch failure to eval any outPath of hydra jobset (with aliases turned off) | 23:15:39 |
emily | (unrelated to the change) | 23:15:41 |
emily | OTOH I'm not sure how ofborg hasn't complained about every PR since? | 23:15:49 |
emily | so maybe it is actually just some accursed mystery | 23:15:56 |
Lily Foster | oh it's probably because of the system fuckery that happens | 23:17:58 |
emily | does ofborg somehow evaluate as all systems simultaneously | 23:21:10 |
Lily Foster | well, it uses the hydra jobset | 23:21:21 |
Lily Foster | so yes | 23:21:24 |
Lily Foster | in a sense | 23:21:32 |
Randy Eckenrode | In reply to @emilazy:matrix.org I think the answer is that it didn't actually finishing evaluating the final version because we rushed before the freeze, and IIRC on previous revisions there was some other eval-breaking thing. I ran the ofborg check locally. I did confirm eval with the SDK rework before I marked it ready. It’s very strange that it failed on Hydra, but I’ve also encountered a case where my tree passed the ofborg check, but something I tried to build failed to eval. | 23:21:48 |
17 Oct 2024 |
Lily Foster | In reply to @emilazy:matrix.org so maybe it is actually just some accursed mystery config.allowUnsupportedSystem = true | 00:00:31 |
emily | ah | 00:00:39 |
emily | not so accursed | 00:00:42 |
emily | but also, come on | 00:00:44 |
Lily Foster | that was what i suspected when i realized the hydra jobset was what was failing to eval (i didn't read context) | 00:00:50 |
Lily Foster | let me double check though | 00:01:02 |
Lily Foster | oh no it's not that at all | 00:02:08 |
Lily Foster | https://github.com/NixOS/ofborg/blob/5a4e743f192fb151915fcbe8789922fa401ecf48/ofborg/src/outpaths.nix#L22-L25 | 00:02:09 |
Lily Foster | but also | 00:02:11 |