| 25 Feb 2026 |
 problems | i know i had this crossSystem toolchain working in 2022 so something changed since then i guess? | 20:21:01 |
 nim65s | maybe these are just issues from a newer gcc ? | 20:23:40 |
| nim65s changed their display name from Guilhem to nim65s. | 20:23:45 |
| problems | it's not even building rust at this point so i guess a minimal repro would just be trying to build for system = "i486-netbsd" | 20:24:00 |
| problems | it's not even building rust at this point so i guess a minimal repro would just be trying to build gcc for system = "i486-netbsd" | 20:24:15 |
| 26 Feb 2026 |
| problems | bisecting nixpkgs now ig | 01:16:06 |
 Lun | I'd guess GCC 13 or 14 bump with it being a Wint-conversion failure | 01:39:56 |
| Lun | yeah 14 changed that https://gcc.gnu.org/gcc-14/porting_to.html#int-conversion | 01:44:54 |
| problems | hm okay maybe just bumping the netbsd sources might work then? if i can find them | 01:57:29 |
| problems | i incremented the version from 9.2 to 9.4, so here's hoping | 02:00:17 |
| problems | nope ;~; | 03:09:33 |
| problems | my netbsd wife is saying that netbsd is only supposed to be built with its own patched gcc so idk :l | 03:15:43 |
|  @somasis:matrix.org left the room. | 04:25:06 |
 szlend | Just wondering, are LLVM version additions usually backported to stable channels? I've seen LLVM 21 backported last time to 25.05, but 22 is missing in 25.11 this time | 11:33:31 |
 K900 | Usually not unless they're necessary for something else | 11:33:51 |
 Alyssa Ross | I would generally backport. It makes it easier to backport other LLVM unstable PRs. But indeed this time I think it didn't happen. Still could. | 11:34:37 |
| szlend | I wonder how big of a lift that is. I guess it depends on how much the stdenv infra has changed in unstable? | 11:37:45 |
| szlend | Rust has started to use LLVM 22 in nightly, so I imagine they're gonna upgrade soon. Which breaks some things like cross boundary lto. But don't think that actualy affects nixpkgs itself | 11:39:20 |
| Alyssa Ross | I don't think it would be difficult. | 11:40:17 |
| szlend | I got a build running by applying a diff between 25.11/unstable for paths llvm/22, llvm/common and the llvm stuff in all-packages.nix. Hopefully that passes | 16:51:50 |
| szlend | Say I get this build passing. To make an acceptable PR I assume you'd want me to look up and backport all the individual llvm-22 and llvm-common commits from unstable right? | 16:55:11 |
| szlend | Though some changes in llvm/common are not directly related to llvm-22 | 16:56:17 |
| szlend | bleh, one llvm test failed (TestExecuteEmptyEnvironment, missing libLLVM.dylib) | 21:15:32 |
 Tristan Ross | There's a fix for that on staging | 21:24:37 |
| Tristan Ross | The PR which fixes it is linked in the LLVM 22 PR | 21:24:49 |
| szlend | ah great | 21:25:08 |
| problems | okay so is it possible to build gcc13Stdenv without a dependency on gcc15? | 23:42:41 |
| problems | it still tries to build libcMinimal with gcc15 apparently | 23:43:21 |
| 27 Feb 2026 |
| Tristan Ross | Huh what, what are you trying to accomplish? | 00:01:43 |
| problems | netbsd stdenv is broken on gcc14 and newer | 00:02:07 |
