| 4 Apr 2026 |
 Randy Eckenrode | That should be fixed for a while though in Nixpkgs. | 01:14:48 |
 antifuchs | so, I was looking for virtualisation.darwin-builder in the nix-darwin and nixos docs, and neither seems to have that documented. how did you find that? | 01:39:43 |
 Ihar Hrachyshka | $ pwd
/Users/ihrachyshka/src/nixpkgs
$ git grep darwin-builder doc | grep diskSize
doc/packages/darwin-builder.section.md: virtualisation.darwin-builder.diskSize = 5120;
| 01:40:49 |
| antifuchs | Ah, hm, maybe that’s just not in the online nix-darwin manual options list | 01:55:23 |
| Randy Eckenrode | https://github.com/NixOS/nixpkgs/pull/506470 fixes the ICU problem on Darwin. | 02:14:50 |
 EsperLily [she/her] | i'm not 100% positive but i think the MacVim.app/Contents/MacOS/Vim binary is in fact the vim binary produced by the classic vim compilation | 03:22:26 |
| EsperLily [she/her] | the working MacVim was built with llvm-21.1.7, the broken one is built with llvm-21.1.8 | 03:23:26 |
| EsperLily [she/her] | * i'm not 100% positive but i think the MacVim.app/Contents/MacOS/Vim binary is in fact the vim binary produced by the classic vim compilation, meaning without Xcode's involvement | 03:33:19 |
| EsperLily [she/her] | yeah the darwin-builder is actually documented in nixpkgs https://nixos.org/manual/nixpkgs/stable/#sec-darwin-builder-reconfiguring | 03:42:49 |
| EsperLily [she/her] | i just confirmed that if I swap out stdenv.cc with 21.1.7 the crash goes away. I declared a new input nixpkgs-old which is pinned to the older nixpkgs that was working fine and added an overlay ```
(final: prev: {
macvim = prev.macvim.override {
stdenv = final.stdenvAdapters.overrideCC final.stdenv nixpkgs-old.legacyPackages.${system}.stdenv.cc;
};
})
| 04:09:55 |
| EsperLily [she/her] | * i just confirmed that if I swap out stdenv.cc with 21.1.7 the crash goes away. I declared a new input nixpkgs-old which is pinned to the older nixpkgs that was working fine and added an overlay
(final: prev: {
macvim = prev.macvim.override {
stdenv = final.stdenvAdapters.overrideCC final.stdenv nixpkgs-old.legacyPackages.${system}.stdenv.cc;
};
})
| 04:10:12 |
| EsperLily [she/her] | * i just confirmed that if I swap out stdenv.cc with 21.1.7 the crash goes away. I declared a new input nixpkgs-old which is pinned to the older nixpkgs that was working fine and added an overlay
(final: prev: {
macvim = prev.macvim.override {
stdenv = final.stdenvAdapters.overrideCC final.stdenv nixpkgs-old.legacyPackages.${system}.stdenv.cc;
};
})
and this one works without crashing
| 04:10:24 |
| EsperLily [she/her] | argh it failed again | 04:29:10 |
| Randy Eckenrode | Building ffmpeg 8 appears to be broken on Darwin. It’s trying to use patchelf ….
> patching script interpreter paths in /nix/store/fvxp6blkrg3h37l348lb4f5aq3v7z67y-ffmpeg-headless-8.0.1
> /nix/store/6ifjywcskbcx5mzy4g1gnygdzgfagfs8-stdenv-darwin/setup: line 267: patchelf: command not found
| 10:53:46 |
| Randy Eckenrode | * Building ffmpeg 8 appears to be broken on Darwin. It’s trying to use patchelf ….
> patching script interpreter paths in /nix/store/fvxp6blkrg3h37l348lb4f5aq3v7z67y-ffmpeg-headless-8.0.1
> /nix/store/6ifjywcskbcx5mzy4g1gnygdzgfagfs8-stdenv-darwin/setup: line 267: patchelf: command not found
| 10:53:53 |
| Randy Eckenrode | Oh, wait. That’s my Vulkan patches. That’s my fault. | 10:54:24 |
| antifuchs | Ah, GitHub actions levels of documentation indirection (: | 11:39:40 |
 emily | https://github.com/llvm/llvm-project/compare/llvmorg-21.1.7...llvmorg-21.1.8 | 11:42:16 |
| emily | either one of these or a packaging change | 11:42:34 |
| emily | best guess... https://github.com/llvm/llvm-project/commit/d2e835bd46384aa1628c9d44e3f93438f0795dfd? | 11:43:10 |
| emily | https://github.com/rust-lang/rust/issues/149522 | 11:43:28 |
| emily | would be wild if this somehow broke vim | 11:43:39 |
| emily | try building without optimization? | 11:44:05 |
|  danieroux set a profile picture. | 13:52:43 |
| EsperLily [she/her] | i just tried, setting CFLAGS = -g -O0 and it's still crashing | 20:16:19 |
| emily | try it with -O0 with the old compiler | 20:16:46 |
| emily | since if it is that commit, it'd be an incorrect optimization in LLVM somehow causing it to not crash | 20:17:01 |
| emily | seems bizarre but who knows | 20:17:10 |
| emily | it could ofc be anything else in the stdenv.cc closure too | 20:17:25 |
| emily | (linker stuff…?) | 20:17:38 |
