| 14 Oct 2025 |
 ivy | now this oops https://github.com/NixOS/nixpkgs/pull/451871 | 05:31:40 |
 Jan Tojnar | any tips why sincos function would suddenly not be present on Darwin? | 16:46:30 |
| Jan Tojnar | The build is still defining _GNU_SOURCE | 16:46:32 |
| Jan Tojnar | regarding https://github.com/NixOS/nixpkgs/pull/449689 | 16:46:41 |
 emily | yes | 16:47:49 |
| emily | the LLVM update added __builtin_sincos or something | 16:47:55 |
| emily | the Meson check checks for this, but the GTK code can't use it | 16:48:00 |
| emily | this came up in the macOS room let me dig up the correct fix | 16:48:08 |
| emily | Jan Tojnar:
https://github.com/mesonbuild/meson/issues/3740
https://matrix.to/#/!lheuhImcToQZYTQTuI:nixos.org/$JJBUjNnTLa7Ii2Dy5f7hCCFcbWlnIawrx4sWNwx1AXo?via=nixos.org&via=matrix.org&via=nixos.dev
upstream should pass e.g. prefix : '#include <math.h>', args: '-D_GNU_SOURCE' to has_function
| 16:49:26 |
| emily | rather than special-casing Clang | 16:49:36 |
| emily | it is arguably a Meson bug to accept __builtin_foo when you asked to check for foo | 16:50:06 |
| emily | IMO | 16:50:13 |
| Jan Tojnar | Thanks, forwarded that. | 17:09:22 |
| 15 Oct 2025 |
 Vladimír Čunát | I anticipate significant trouble with bmake on darwin in the current staging-next. A test is failing reliably there now, and it blocks important packages (including the channel). But changing bmake will normally rebuild darwin stdenv... | 10:36:53 |
| Vladimír Čunát | (the build during stdenv bootstrap did pass but the later one does not) | 10:37:52 |
 Randy Eckenrode | bmake is built only once during the bootstrap. You can detect the stage from the stdenv name to do a staging-next fix that avoids a rebuild then drop the condition in staging. | 11:20:55 |
| Randy Eckenrode | There are a couple of open issues about it. Meson devs might take patches, but they’re not actively working on a fix. | 11:22:24 |
| Randy Eckenrode | Darwin has sincos, but it’s prefixed with an underscore. I’m not sure how the _GNU_SOURCE thing solves the problem on Darwin, but if it does …. | 11:23:39 |
| Vladimír Čunát | Do you have a reference at hand for how to best do this detection? | 11:23:53 |
| Randy Eckenrode | (I fixed this on my working branch by defining sincos to __builtin_sincos.) | 11:24:13 |
| Vladimír Čunát | (I'd probably do at least quick bmake fixup to try at least building the blockers.) | 11:24:32 |
| Vladimír Čunát | * (I'd probably do at least quick bmake fixup to try at least building the blockers depending on it.) | 11:24:37 |
| Randy Eckenrode | https://github.com/NixOS/nixpkgs/blob/master/pkgs/development/compilers/llvm/common/llvm/default.nix#L49 | 11:25:43 |
| Randy Eckenrode | Except it would be stage1 not stage-xclang. | 11:27:06 |
| Vladimír Čunát | Oh right, I could've used lib.trace to find the names. | 11:37:00 |
| Randy Eckenrode | Once the LLVM stuff lands that makes overriding work, it should be possible to move that check to an override in the bootstrap. | 11:38:41 |
 dramforever | maybe it's os dependent? | 11:39:16 |
| dramforever | what if it just fails on intense-heron but works on maximum-snail | 11:39:43 |
| Vladimír Čunát | The bmake issue is triggered by existence of some file, if I googled fast right. We've encountered this one already, apparently. | 11:41:11 |
| dramforever | i saw some stuff that say that but i can't reproduce | 11:42:40 |
