| 20 Nov 2025 |
Tristan Ross | Now, this isn't working as expected but it does eval at least | 03:45:32 |
Tristan Ross | https://github.com/NixOS/nixpkgs/pull/463361 | 03:45:41 |
| John joined the room. | 05:47:25 |
Grimmauld (any/all) | In reply to @rosscomputerguy:matrix.org Starting to work on native LLVM stuff Nice! Then I'll not have to do cursed same-triple-cross for musl llvm! | 08:07:52 |
bake.monorail | What are you using to get this CLI output? | 08:08:57 |
Tristan Ross | In reply to @grimmauld:grapevine.grimmauld.de Nice! Then I'll not have to do cursed same-triple-cross for musl llvm! Heh yeah and it's nice that it bootstraps via the existing GCC tarball | 08:09:06 |
Tristan Ross | In reply to @bake.monorail:matrix.org What are you using to get this CLI output? nom | 08:09:10 |
Grimmauld (any/all) | All the Qt stack is cursed enough on regular cross. It completely breaks when the target is executable on the build bit cmake still thinks its cross. | 08:10:10 |
Tristan Ross | In reply to @grimmauld:grapevine.grimmauld.de All the Qt stack is cursed enough on regular cross. It completely breaks when the target is executable on the build bit cmake still thinks its cross. Same for gfortran | 08:11:25 |
Tristan Ross | So having a DE using LLVM cross is just not really practical without disabling a few dozen things | 08:12:05 |
Tristan Ross | I'm hoping we can get this into 26.05 or 26.11, would be awesome | 08:12:49 |
Grimmauld (any/all) | I have sway and quickshell building on same triple cross musl llvm, and the VM starts | 08:12:54 |
Grimmauld (any/all) | Takes like fifty overlays, but it doss work | 08:13:16 |
Tristan Ross | The problem is audio stuff lol | 08:13:23 |
Grimmauld (any/all) | In reply to @rosscomputerguy:matrix.org The problem is audio stuff lol Nah I have audio too | 08:13:32 |
Grimmauld (any/all) | Pipewire | 08:13:35 |
Tristan Ross | In reply to @grimmauld:grapevine.grimmauld.de Nah I have audio too fftw built? | 08:13:51 |
Grimmauld (any/all) | You can just null out gfortran on fftw, it'll use the precompiled FORTRAN bindings then | 08:13:55 |
Tristan Ross | Oh | 08:14:03 |
Grimmauld (any/all) | :p | 08:14:13 |
Grimmauld (any/all) | I'd like to build fftw with flang, but flang is broken | 08:14:28 |
Tristan Ross | Lol | 08:14:38 |
Grimmauld (any/all) | If we do llvm, flang would make sense. Not sure whether it is broken completely, broken because musl, because lld, or broken because llvm, but its broken | 08:15:43 |
Grimmauld (any/all) | https://termbin.com/cnx3 there, maybe you have a better idea than i have about what is wrong here Its probably musl stuff... | 08:35:33 |
Grimmauld (any/all) | i guess i could overlay a glibc flang on the build platform only, that could work... | 08:36:23 |
Tristan Ross | In reply to @grimmauld:grapevine.grimmauld.de https://termbin.com/cnx3 there, maybe you have a better idea than i have about what is wrong here Its probably musl stuff... Yeah, I'm confident to say that's likely the case | 08:44:11 |
Grimmauld (any/all) | https://github.com/NixOS/nixpkgs/pull/461277#issuecomment-3556593423
never mind! I guess system unit generator on musl is a bit wonky right now! So no i currently don't have a musl llvm vm, be back after a systemd recompile | 08:45:30 |
Tristan Ross | Oh it turns out I got a disallowed references error lol. Well that should be easy enough to fix. | 15:43:35 |
Grimmauld (any/all) | openssl? | 17:02:09 |
Grimmauld (any/all) | https://github.com/openssl/openssl/pull/29176
openssl is currently leaking its build compiler into the runtime closure on llvm-based cross and our asserts break, so you might have to fetch this depending on what you build | 17:03:07 |