| 18 Apr 2026 |
 Emil Thorsøe | not insurmountable though | 15:15:47 |
 K900 | There's a whole bunch of issues really | 15:16:15 |
 vcunat | Uh, the autoconf update is such a PITA for darwin. | 21:53:46 |
| vcunat | Too many packages break from getting autoreconf-ed with -std=gnu23 | 21:54:04 |
| vcunat | Many packages just aren't ready for such new stuff. | 21:54:33 |
| vcunat | I hate that C standard even does get changed by autoreconfig. It doesn't make sense. Only upstream should switch that. | 21:57:05 |
| vcunat | * I hate that C standard even does get changed by autoreconf-ing. It doesn't make sense. Only upstream should switch that. | 21:57:23 |
| 19 Apr 2026 |
 Alyssa Ross | Only upstream is supposed to run autoreconf | 06:16:08 |
| Alyssa Ross | That's the design. | 06:16:11 |
| vcunat | I know, but not everyone agrees with that. And sometimes you need to apply patches which may make it hard to avoid it. | 07:57:20 |
| Alyssa Ross | Yep, and if we do that, we're opting into keeping up with autoconf changes and should be taking that into consideration. | 08:01:35 |
| Alyssa Ross | It's also changed by new compiler versions by default, right? So even without autoconf it's not solely on upstream. | 08:05:44 |
| Alyssa Ross | What we can do is make sure every time we have to pin a C standard, we send that change upstream so going forward upstream will take responsibility for it | 08:07:37 |
| vcunat | We could. But that's even increasing the amount of work that we need to do. | 08:10:57 |
| vcunat | Like, I'm not really motivated to spend many hours on this, especially as it doesn't even affect Linux. | 08:12:03 |
| vcunat | * Like, I'm not really motivated to spend many hours on this, especially as it doesn't even affect Linux (currently). | 08:12:12 |
| Alyssa Ross | Sure, not saying you should. | 08:12:31 |
| Alyssa Ross | We have the big darwin-maintainers team for issues like this, I think | 08:12:47 |
| vcunat | But I don't want to block staging-next either or merge it with very large regressions. | 08:12:48 |
| Alyssa Ross | Our contributor base is quite good at mass stuff like this IME if they're given clear instructions. | 08:13:38 |
| Alyssa Ross | If you link me an example on Hydra I could write up an issue. | 08:14:03 |
| Alyssa Ross | The burden of this sort of thing shouldn't just be on a few people doing everything | 08:15:17 |
| Alyssa Ross | Why is it only for macOS? LLVM update? | 08:20:22 |
| Alyssa Ross | * | 08:21:10 |
| vcunat | It's possible that it's about clang vs. gcc. | 08:29:27 |
| Alyssa Ross | The autoconf release notes say it uses c23 "when available" | 08:29:39 |
| Alyssa Ross | so I'd guess that for some reason it's newly available for Darwin specifically, which probably means Clang specifically? | 08:29:59 |
| Alyssa Ross | Yeah I can reproduce one of them from Hydra with pkgsLLVM | 08:30:15 |
| Alyssa Ross | That's also good — macOS is not necessarily required to fix | 08:31:11 |
| vcunat | Examples I saw now (because many jobs depend on them):
https://hydra.nixos.org/build/326635250
https://hydra.nixos.org/build/326219525 | 08:32:37 |
