| 14 Nov 2025 |
 K900 | Probably should yeah | 08:27:05 |
 Grimmauld (migrated to @grimmauld:m.grimmauld.de) | There is also something wrong with the cross check:
https://github.com/NixOS/nixpkgs/blob/0d9bcf4d56844168da8ea9d32b01eed507bb0371/pkgs/development/libraries/qt-6/modules/qtbase/default.nix#L90
it is a canExecute check, but this fails on cross targets that can actually be executed, such as building x86_64-unknown-linux-musl as cross (as in, useLLVM = true in the host platform config while on a glibc build system). This makes cmake confused and complain about a missing DQT_HOST_PATH.
Can be fixed by adding (lib.cmakeBool "CMAKE_CROSSCOMPILING" isCrossBuild) cmake flag to force cmake to treat executable cross as native build. | 08:27:11 |
| Grimmauld (migrated to @grimmauld:m.grimmauld.de) | Btw, is this intended:
https://github.com/NixOS/nixpkgs/blob/0d9bcf4d56844168da8ea9d32b01eed507bb0371/pkgs/development/libraries/qt-6/default.nix#L65-L68
https://github.com/NixOS/nixpkgs/blob/0d9bcf4d56844168da8ea9d32b01eed507bb0371/pkgs/development/libraries/qt-6/modules/qtbase/default.nix#L78
withGtk3 defaults to false but is then forced to true. Feels wonky. Shouldn't the default itself be !withMingw if that is really what we want? | 08:27:22 |
| K900 | Qt cross is still extremely cursed | 08:27:43 |
| Grimmauld (migrated to @grimmauld:m.grimmauld.de) | yes | 08:27:48 |
| Grimmauld (migrated to @grimmauld:m.grimmauld.de) | but i am doing it and i think this is a fix we should do | 08:28:00 |
| Grimmauld (migrated to @grimmauld:m.grimmauld.de) | (if it works, we'll see) | 08:28:08 |
| Grimmauld (migrated to @grimmauld:m.grimmauld.de) | feels weird to have executable cross fail while non-executable cross just works | 08:29:08 |
| Grimmauld (migrated to @grimmauld:m.grimmauld.de) | though the issue is the same in other qt components | 08:30:11 |
| Grimmauld (migrated to @grimmauld:m.grimmauld.de) | its a bit painful | 08:30:14 |
| Grimmauld (migrated to @grimmauld:m.grimmauld.de) | * though the issue is the same in other qt components too | 08:30:27 |
| K900 | Any effort towards unfucking it will be greatly appreciated | 08:31:07 |
| Grimmauld (migrated to @grimmauld:m.grimmauld.de) | the proper fix is probably some cmake fuckery... | 08:33:20 |
| K900 | I think the proper fix is to basically build a "cross qt" properly, with targetPlatform | 08:36:24 |
| K900 | Which is very cursed | 08:36:30 |
| Grimmauld (migrated to @grimmauld:m.grimmauld.de) | the problem is, the canExecute check is different than whatever cmake checks to detect cross, and that mismatch causes issues | 08:42:53 |
