| 19 Nov 2025 |
 Randy Eckenrode | If I build against release LLVM, mesa_clc crashes instead. | 18:55:43 |
 Lun | RelWithDebug LLVM and a debugger might be a better starting point with things like https://github.com/llvm/llvm-project/issues/71174 lying around to hit in debug builds | 19:06:10 |
| Lun | * relwithdebinfo LLVM and a debugger might be a better starting point with things like https://github.com/llvm/llvm-project/issues/71174 lying around to hit in debug builds | 19:06:20 |
| Lun | actually that error sounds familiar from fuckups with multiple different llvm versions linked into the same process | 19:08:23 |
| Randy Eckenrode | I got overrideScope working correctly, so everything should have a consistent LLVM. | 19:10:30 |
| Lun | iirc rocMLIR + rocm-comgr in the same process would trigger LLVM ERROR: Option 'default' already exists! followed by an abort if the llvm versions they linked against weren't matching | 19:10:35 |
| Randy Eckenrode | mesa_clc links against SPIRV-LLVM-Translator, which is linked against the debug build of libLLVM.dylib. | 19:11:50 |
|  captain_love joined the room. | 19:14:23 |
| Randy Eckenrode | Redacted or Malformed Event | 20:23:19 |
| Lun | Any luck figuring it out? | 22:05:01 |
| Randy Eckenrode | No, but I haven’t been able to dig into it for a bit. Release with debug symbols wasn’t enough, unfortunately. | 22:07:02 |
| 20 Nov 2025 |
 EsperLily [she/her] | emily reviewed my macvim PR but didn't merge it (and then i made updates to it based on her comments). she doesn't seem to have been active in the past couple of weeks, could i get someone else to merge it? https://github.com/NixOS/nixpkgs/pull/455760 | 00:34:30 |
| Randy Eckenrode | Strictly speaking, the 14.4 SDK isn’t required anymore. It’s the default now on unstable. I don’t want to delay it again just to get that fixed, so I merged. | 00:40:33 |
| EsperLily [she/her] | thanks! I'll have to remember that the next time there's an update to be made | 00:41:39 |
| Randy Eckenrode | KosmicKrisp builds if I use the upstream meson options. Now I need to figure out what’s wrong with the ones being passed from nixpkgs. | 03:24:33 |
|  John joined the room. | 04:44:16 |
 K900 | Doesn't the Darwin expression have like, uh, two of them | 07:21:55 |
| Randy Eckenrode | It needs something else, or mesa_clc comes out broken. Upstream relies on autodetection, but that’s disabled in the expression, so I have to figure out what and enable it. | 11:33:13 |
| K900 | Huh | 11:33:42 |
| K900 | OK | 11:33:50 |
| K900 | Feel free to reenable autoFeatures if you think it helps | 11:33:59 |
| K900 | It's mostly disabled on Darwin so it builds as little as possible | 11:34:12 |
| Randy Eckenrode | Is that why Darwin is separate? Should I reenable Darwin back in the main derivation instead of working on the Darwin one? | 12:20:37 |
| K900 | Uhhh | 12:20:55 |
| K900 | Doubt it tbh | 12:20:57 |
| K900 | Because there's actually not a lot shared | 12:21:04 |
| K900 | Like the Darwin derivation is separate because it's basically not building the same thing at all right now | 12:21:18 |
| K900 | There will probably be a little more overlap with Vulkan now | 12:21:33 |
| K900 | But basically none of the Linux stuff applies to Darwin | 12:21:50 |
| Randy Eckenrode | I’ll need to look at it again tonight, but enabling Vulkan for KosmicKrisp makes it more similar IIRC. | 13:00:32 |
