| 27 Apr 2026 |
 hexa (signing key rotation when) | sterni? | 11:16:54 |
 sterni | beautiful | 11:17:44 |
| sterni | isn’t the timeout 3h though or just if there’s no log output? | 11:18:03 |
| hexa (signing key rotation when) | no log | 11:18:09 |
| hexa (signing key rotation when) | 2h on darwin, 3h on linux | 11:19:13 |
| hexa (signing key rotation when) | we could add a global timeout, but | 11:19:21 |
| hexa (signing key rotation when) | at 24h or so | 11:19:31 |
| sterni | ah test suite that never passes incredible | 11:19:47 |
| hexa (signing key rotation when) | we generally notice extremly long running builds and kill/poke | 11:20:02 |
| hexa (signing key rotation when) | they are fairly rare, so it has not been a huge problem | 11:20:19 |
| hexa (signing key rotation when) | and having some visiblity is better than ignoring them | 11:20:36 |
|  @corngood:corngood.com left the room. | 11:23:34 |
 mynacol | Were all the builder very busy the last hours? All the rusty-v8 builds for deno 2.7.13 failed, almost all on timeouts. At that build point in the logs, the major compilation work happens without any log output, but that should only take 30 min at max (more like 20 min IME). Maybe restart these jobs and/or increase the timeout for that derivation?
- https://hydra.nixos.org/build/327593597
- https://hydra.nixos.org/build/327607910
- https://hydra.nixos.org/build/327607913
- https://hydra.nixos.org/build/327607911
| 14:22:18 |
| hexa (signing key rotation when) | they all get stuck at temporal_capi | 14:23:36 |
| hexa (signing key rotation when) | so I suspect silent timeout | 14:23:42 |
| hexa (signing key rotation when) | ^ | 14:23:49 |
| hexa (signing key rotation when) | are our builders busy? yes, generally they are. | 14:24:06 |
 leona | and they run not on big-parallel, that means very few cores. Compile likely takes longer than you think. | 14:24:15 |
| hexa (signing key rotation when) | normal builds get 2 cores | 14:24:20 |
| mynacol | Ahh, that probably explains it. Shall I move it to big-parallel? (it's still smaller than firefox/chromium) | 14:25:08 |
| mynacol | * Ahh, that probably explains it. Shall I move it to big-parallel? (it's still a lot smaller than firefox/chromium) | 14:25:19 |
| hexa (signing key rotation when) | right now 59 packages have the big-parallel flag 🤔 | 14:26:34 |
 vcunat | I'd say that if build without the flag takes hours and the flag speeds it up very significantly, we'd better add the flag. | 14:28:25 |
| vcunat | * I'd say that if a build without the flag takes hours and the flag speeds it up very significantly, we'd better add the flag. | 14:28:31 |
| vcunat | * I'd say that when a build without the flag takes hours and the flag speeds it up very significantly, we'd better add the flag. | 14:28:39 |
| mynacol | Or just increase the the timeout. (My compiles with 32 core somewhat older intel take 30 min for rusty-v8 and 1h for deno, and for comparison 1-2h (depending on pgo double-runs) on mullvad-browser) | 14:28:53 |
| mynacol | * Or just increase the timeout. (My compiles with 32 core ~~somewhat older~~ intel Xeon E5-2630 v3 take 30 min for rusty-v8 and 1h for deno, and for comparison 1-2h (depending on pgo double-runs) on mullvad-browser) | 14:32:20 |
|  Ninja joined the room. | 14:37:10 |
| mynacol | The funny thing is that the builds are at those points rather single-threaded (only one or two hefty crates). So I look into the timeout instead: https://github.com/NixOS/nixpkgs/pull/514057 | 14:51:28 |
| 28 Apr 2026 |
|  Aangularity joined the room. | 04:39:30 |
