| 28 May 2026 |
 K900 | I bet like 90% of that stuff doesn't even need xvfb | 09:39:52 |
| K900 | Because it doesn't even use X on Darwin | 09:40:00 |
| K900 | Sigh | 09:40:10 |
 Randy Eckenrode | The build process hardcodes it as .so. I don’t know what X expects or even if it uses that. That’s for llvmpipe, isn’t it? | 11:27:22 |
| Randy Eckenrode | It doesn’t affect KosmicKrisp. If we weren’t also building GLX support, we wouldn’t have even noticed. | 11:28:53 |
| Randy Eckenrode | Feature-wise, KosmicKrisp should be close to MoltenVK or exceed it by 26.2. There’s been a lot pf development over the last few weeks. Tessellation should be landing soonish. | 11:30:29 |
| Randy Eckenrode | * | 11:30:51 |
| Randy Eckenrode | Performance is less certain. It’s not been a focus so far. | 11:31:38 |
| K900 | Yeah | 11:35:13 |
| K900 | The build process does, but where does X look for it | 11:35:25 |
| K900 | Is the question | 11:35:27 |
| K900 | Like is the mistake that it's .dylib in the first place | 11:35:34 |
| K900 | Or is the mistake that the script can't handle .so | 11:35:45 |
| Randy Eckenrode | They changed the script years ago to be platform-independent, but Mesa only started using it on Darwin with 26.1. I don’t know what changed to make it do that. | 11:36:36 |
| Randy Eckenrode | (Because I was too lazy to bisect it.) | 11:36:46 |
| K900 | Presumably dril | 11:37:32 |
 viraptor | In reply to @kfiz:matrix.org
https://github.com/NixOS/nixpkgs/issues/525087 Yeah, that's on me. Another case to test against... Fix incoming. | 14:23:46 |
| 29 May 2026 |
 Myria | Hey, I was building staging on darwin and noticed that the recent darwin update cause the darwin build to fail (both x86 and arm). I've got a quick patch that disables the failing test, but is there something I should do to assure that this isn't a problem with upstream rsync and just our environment we build it in? | 06:53:52 |
| Myria | Hmm, seems like homebrew has pushed the latest version, so it's probably just a nixpkgs issue | 06:54:58 |
| Myria | * | 07:03:53 |
| Myria | *rsync | 07:03:56 |
| Myria | Rsync update not darwin update. That was an oops on my part | 07:04:11 |
| viraptor | Rsync had some badness in the latest release, I've seen at least 3 people run into that one already today in unrelated ways.. | 07:14:04 |
| Myria | Yeah, fair enough. I've got a fix I'm just testing now that disables the failing tests on darwin, and hopefully that should fix things. | 07:20:23 |
|  kfiz set a profile picture. | 23:10:45 |
| 30 May 2026 |
| viraptor | Got a fix for you: https://github.com/NixOS/nixpkgs/pull/525811 | 01:54:10 |
| viraptor | * Got a fix for you. There was a lot of new weird stuff in macos26 icon generation: https://github.com/NixOS/nixpkgs/pull/525811 | 01:55:03 |
 kfiz | Thx. Works for me. Had to reboot though. When I was directly opening it with open result/Applications/Element.app my machine became unresponsive. After a reboot the app opened normally as part of loginwindow, i.e. as part ot the previously opened applications. | 03:05:33 |
| viraptor | In reply to @kfiz:matrix.sopado.de
Thx. Works for me. Had to reboot though. When I was directly opening it with open result/Applications/Element.app my machine became unresponsive. After a reboot the app opened normally as part of loginwindow, i.e. as part ot the previously opened applications. The icons are parsed by the system and apparently I can crash one of those services. Don't think about it too much. Definitely no security implications there... | 04:48:11 |
| viraptor | * The icons are parsed by the system and apparently I can crash one of those services. (That's carried over from the previous bad file) Don't think about it too much. Definitely no security implications there... | 04:50:57 |
