| 27 Jun 2025 |
 vcunat | Say, firefox building for 5 hours... it is a lot. | 14:59:09 |
| vcunat | Even my cheap old x86 Ryzen it's way faster. | 14:59:44 |
| vcunat | * Say, firefox building for 5 hours... it is a lot. | 15:00:36 |
 hexa | that's wild | 15:18:10 |
| hexa | but it also builds 4 things in parallel usually | 15:18:16 |
| hexa | and the many firefoxes, electrons and chromiums are regularly the stragglers | 15:18:28 |
| hexa | the builds regularly take 90-120m for me | 15:20:19 |
| vcunat | Oh, I confused machines, and the one I use isn't as cheap+old as I thought, but it's still surprising to me. | 15:23:51 |
| vcunat | The long build times (5-10h) are also problematic when we push changes to the Hydra server. When the queue runner gets restarted, hours of this work get lost. | 15:25:50 |
| vcunat | * The long build times (5-10h) are also sometimes problematic when we push changes to the Hydra server. When the queue runner gets restarted, hours of this work get lost. | 15:26:39 |
|  nbp — PTO changed their display name from nbp to nbp — PTO. | 17:25:59 |
| vcunat | Actually, I have no idea what timeout is triggering. | 20:56:14 |
| vcunat | Normal one is surely longer than this (10h IIRC). And max-silent-time is 2h, which did not happen, I hope 🤔 | 20:57:35 |
| hexa | meta.timeout should be 7h | 21:00:00 |
| hexa | * meta.timeout should be 2h | 21:00:09 |
| hexa | * meta.maxSilent is 14400 for firefox | 21:00:35 |
| hexa | so 4h | 21:00:40 |
| hexa | unlikely that it hit that | 21:00:51 |
| vcunat | Or maybe the build really does get stuck and that's why it times out after 2h. | 21:06:03 |
| 28 Jun 2025 |
| hexa | firefox 140.0.1 | 00:51:06 |
| hexa | built for 24.11 | 00:51:11 |
| hexa | takes just under 2h49m to build and complete the nixos test | 00:51:30 |
| hexa | years ago when I started maintaining it I could build firefox in under an hour easily | 00:51:56 |
| hexa | like 50m on my desktop | 00:52:07 |
| hexa | pgo made that worse, but its was worth it for the performance boost | 00:52:23 |
| hexa | so 2x 50m + 15m profiling, just shy of two hours | 00:52:56 |
| hexa | but now we're closer to three hours 🫠 | 00:53:04 |
| hexa | so indeed 4h49m sounds like 2h49 + 2h silent timeout 😬 | 13:11:10 |
| hexa | wondering if bumping that up from 2 to 3 hours would make it work | 13:11:27 |
| hexa | updated the linux builders | 13:18:12 |
