| 18 Oct 2025 |
 dramforever | you seem very vehemently against the idea of nixpkgs just not disabling valgrind for mesa on musl+llvm | 12:33:56 |
 matthewcroughan | No, I just think it's probably the right thing to do, like who needs valgrind on mesa? | 12:34:25 |
| dramforever | this is the "urgency" i've been talking about, you've been urging this for like twenty minutes | 12:34:31 |
| matthewcroughan | Like, it's for development and debugging, no? | 12:34:37 |
| matthewcroughan | and it increases the closure size quite a bit probably anyway | 12:34:54 |
| matthewcroughan | Ah yeah it also depends on perl | 12:35:33 |
| matthewcroughan | So can't have a perlless system with this in :D | 12:35:46 |
| dramforever | it is for development and debugging, which a lot of nixpkgs users do | 12:35:50 |
 Alyssa Ross | if you want to argue for disabling valgrind by default on mesa, that's very different from arguing for doing it on "some platforms". | 12:36:25 |
| Alyssa Ross | but presumably at some point somebody decided that Valgrind should be included in Mesa, so it's not necessarily going to be easy to find consensus for | 12:37:26 |
| matthewcroughan | It's optional via override so it's fine | 12:37:46 |
| matthewcroughan | Like I think if it's optional via override, then it doesn't need to touch nixpkgs | 12:38:23 |
| matthewcroughan | And if it's not optional via override like this is, then that's the thing that needs to change. But things should also compile with musl without override, which in this case would require fixing valgrind upstream | 12:39:12 |
| Alyssa Ross | again, mesa works fine with musl | 12:39:26 |
| matthewcroughan | s/musl/llvm/g | 12:39:40 |
| matthewcroughan | I just woke up :D | 12:39:50 |
| matthewcroughan | Crawled to my computer and started building | 12:40:02 |
| Alyssa Ross | support for building with LLVM in Nixpkgs is very new. musl has had a lot of work put into it over years, and LLVM hasn't yet. | 12:40:12 |
| Alyssa Ross | (speaking from experience) it's also very important when doing stuff with obscure platforms to make sure you're not stepping on anybody's toes, because if people get annoyed they'll start wanting to just declare your platform unsupported, and then you're screwed | 12:40:23 |
| Alyssa Ross | the #1 most important thing for musl support in Nixpkgs being where it is today (where I don't have to have a fork) is that it happened quietly, so nobody ever got annoyed enough to ask why they should have to care about it | 12:41:56 |
| matthewcroughan | Who merged the PRs and reviewed the PRs though? | 12:42:31 |
| Alyssa Ross | package maintainers mostly | 12:42:54 |
| Alyssa Ross | I always tried to make very sure that PRs were merge-ready when I opened them | 12:43:08 |
| Alyssa Ross | in many cases, there were no Nixpkgs PRs | 12:43:40 |
| Alyssa Ross | because I got things fixed upstream | 12:43:44 |
| Alyssa Ross | and just waited for Nixpkgs to update | 12:43:50 |
| matthewcroughan | I see, yeah so like I said if I find any more things like that I'll just make it a draft PR instead of an open one | 12:44:12 |
| Alyssa Ross | I tried to make it so that other people had to spend as little time thinking about musl as possible | 12:44:40 |
| Alyssa Ross | (this did mean a lot of suffering in silence, because at the time there wasn't really anybody else around who cared — nowadays fortunately that's not true and we can help each other.) | 12:46:11 |
| Alyssa Ross | often I just had to put some stuff down until I had acquired more experience with simpler fixes and could come back to it later | 12:46:46 |
