| 12 Dec 2025 |
 K900 | Or try https://github.com/NixOS/nixpkgs/commit/8ae30516b638d364cbc121fe307e841a9f22f737 | 12:09:35 |
| K900 | (it'll be a kitchen sink rebuild though) | 12:09:42 |
 aktaboot (nvidia, 🖕 u) | and i'd preferrably set all my system to that branch for testing right | 12:10:56 |
| K900 | Well | 12:11:25 |
| K900 | You can overlay qtdeclarative with just that patch | 12:11:30 |
| K900 | It'll be fine | 12:11:33 |
| K900 | But you'd still have to rebuild all of Qt basically | 12:11:49 |
| K900 | Including qtwebengine which is a chromium | 12:11:54 |
| aktaboot (nvidia, 🖕 u) | okay might not try that then :> | 12:12:10 |
| aktaboot (nvidia, 🖕 u) | two weeks is fine | 12:12:37 |
| K900 | Yeah I think this is going to be like | 12:12:40 |
| K900 | Double digit hours on reasonable hardware | 12:12:49 |
| K900 | Or like 3-4 hours of Big Computer | 12:13:10 |
| aktaboot (nvidia, 🖕 u) | and triple digits on a laptop :D | 12:13:27 |
| K900 | Nah | 12:13:34 |
| K900 | Definitely not that bad | 12:13:38 |
| aktaboot (nvidia, 🖕 u) | yeah im jokin | 12:13:40 |
| K900 | qtwebengine is like a 6 hour build on a normal Hydra build slot and those are both tiny and overcommitted to hell | 12:14:05 |
| K900 | @raboof so wait does this mean there's no inherent issue | 12:23:42 |
| K900 | And it's just a link order thing | 12:23:45 |
 raboof | no it's not just link ordering: when dependencies are not explicit, cmake will kick off the qml code generation for the 'dependent' before processing the 'dependee' has finished, causing it to nondeterministically miss generating code that should come in via the 'dependee'. I'm not sure why that doesn't seem to cause more actually-observed issues, though. | 12:27:36 |
| K900 | Wew | 12:27:48 |
| K900 | So it is ordering | 12:27:52 |
| K900 | But yeah | 12:27:54 |
| raboof | * no it's not just link ordering: when dependencies are not explicit, cmake will kick off the qml code generation for the 'depending' before processing the 'dependee' has finished, causing it to nondeterministically miss generating code that should come in via the 'dependee'. I'm not sure why that doesn't seem to cause more actually-observed issues, though. | 12:27:57 |
| K900 | That means we have to fix them one by one then | 12:28:01 |
| K900 | Is really what I meant | 12:28:07 |
| K900 | This sucks | 12:28:10 |
| K900 | (and it doesn't cause issues presumably because the cache is entirely optional and it will just JIT if something isn't cached) | 12:28:33 |
| raboof | right, ordering in time more than ordering in order. but yeah, each will need individual fixes sadly. in a perfect world of course the generator would somehow pick up on such missing dependencies, but I'm already in way over my head understanding as much/little of that build system as I do :) | 12:29:04 |
