| 28 Jul 2025 |
 jade_ | https://gerrit.lix.systems/c/lix/+/3790 code | 06:06:10 |
 K900 | This is so fucking cool | 06:07:35 |
| jade_ | i forget how i came up with this idea, i think it was literally just i dreamt it up while sleep deprived lmao | 06:08:35 |
 aloisw | In reply to @raitobezarius:matrix.org
we should maybe get you a Darwin remote builder if you're interested? If it doesn't cause too much trouble or trust issues, this might be great for iterating on the prelinking. | 06:43:10 |
| jade_ | do you have access to the nix-community machine? | 06:48:10 |
| aloisw | No. | 06:48:51 |
| jade_ | though that machine is uhhh, well abused, so if we have machines to give you access to that would almost definitely be better. raito has more context on that. | 06:49:21 |
| jade_ | i also think that as a separate stream of things, in addition to offering a community shared machine, it would be good to also offer sending devs physical macOS machines if they need them, because base mac minis have gotten to the point of this being realistic. | 06:52:16 |
 helle (just a stray cat girl) | right, I am going to try and get back in some of the dev loop today or tomorrow | 11:16:30 |
 Puna | builtins.currentSystem from pkgsi686Linux.lix gives x86-linux which is nonesense, so I guess the fixups are completely unused, but supposed to be used. submitted separately since this affects more platforms than just POWER: https://gerrit.lix.systems/c/lix/+/3793 | 12:24:21 |
| Puna | * i think I submitted a theoretical fix for the POWER names? not super familiar with the gerrit workflow yet.
https://gerrit.lix.systems/c/lix/+/3773
if host_cpu is supposed to be used there, should i make that a separate changelist thingy? | 12:25:07 |
 Sergei Zimmerman (xokdvium) | In reply to @opna2608:matrix.org
builtins.currentSystem from pkgsi686Linux.lix gives x86-linux which is nonesense, so I guess the fixups are completely unused, but supposed to be used. submitted separately since this affects more platforms than just POWER: https://gerrit.lix.systems/c/lix/+/3793 That has also been broken in cppnix and recently fixed. You’d also need to fix up MIPS/32 bit Arm. | 12:34:34 |
| Sergei Zimmerman (xokdvium) | I think I’ve squashed all problems with that. It doesn’t help tests that there don’t seem to be good integration tests for that stuff. | 12:36:25 |
| Sergei Zimmerman (xokdvium) | * I think I’ve squashed all problems with that. It doesn’t help that there don’t seem to be good integration tests for that stuff. | 12:36:50 |
| Puna | 32-bit ARM should already be handled: https://git.lix.systems/lix-project/lix/src/commit/ca12657a6805045a7e81b2ea761e1af66c4ea1a7/meson.build#L209-L210
MIPS would need added handling though, yes. | 12:37:11 |
| Puna | …oh wait, it's just arm in meson. welp | 12:37:44 |
| Puna | i guess i'll add some checks for those as well then | 12:38:42 |
| Sergei Zimmerman (xokdvium) | In reply to @opna2608:matrix.org
32-bit ARM should already be handled: https://git.lix.systems/lix-project/lix/src/commit/ca12657a6805045a7e81b2ea761e1af66c4ea1a7/meson.build#L209-L210
MIPS would need added handling though, yes. No, I don’t think so. All 32 bit arm targets have arm cpu_family. | 12:46:08 |
| Sergei Zimmerman (xokdvium) | At least that’s what meson docs recommend and that’s what nixpkgs does. | 12:46:40 |
| Puna | added handling for 32-bit ARM & MIPSes to 3793. POWER i was already handling in a separate CL | 13:05:11 |
| Puna | i think that should put it on parity with cppnix then. there's technically still little-endian MicroBlaze which would need to be microblazeel, but i've never heard of that arch before… prolly fine to just leave it unhandled until someone actually needs it?
https://github.com/NixOS/nixpkgs/blob/52c24f184dc4f2980f29e7821063998505d48d56/lib/systems/doubles.nix#L42-L43 | 13:10:50 |
| Sergei Zimmerman (xokdvium) | In reply to @opna2608:matrix.org
i think that should put it on parity with cppnix then. there's technically still little-endian MicroBlaze which would need to be microblazeel, but i've never heard of that arch before… prolly fine to just leave it unhandled until someone actually needs it?
https://github.com/NixOS/nixpkgs/blob/52c24f184dc4f2980f29e7821063998505d48d56/lib/systems/doubles.nix#L42-L43 Makes sense, I’ve only fixed the stuff I could cross-compile and actually test under qemu-user. | 13:26:18 |
 emily | Alyssa is MicroBlaze's strongest soldier :P | 13:29:14 |
| emily | (but I guess not for running it natively) | 13:29:20 |
 Qyriad | qemu-user is super broken — have you actually been able to compile Lix under it? | 13:29:31 |
| emily | you can build a bunch of Nixpkgs stuff with qemu-user | 13:30:22 |
| emily | it does have a lot of rough edges though | 13:30:25 |
| Qyriad | yeah I've compiled a bunch of stuff with it, but I never got Lix (or Nix) to — that was a while ago though | 13:32:11 |
| emily | the actual build or just tests? | 13:39:35 |
| emily | tests breaking doesn't surprise me, the build would | 13:39:39 |
