| 19 Feb 2026 |
 emily | because withWayland ? lib.meta.availableOn stdenv.hostPlatform wayland, is now triggering | 12:00:07 |
| emily | we could presumably add an additional guard there but I am inclined to revert that PR for now because I'm not convinced we should be carrying a somewhat invasive downstream patch to libwayland anyway | 12:01:52 |
| emily | well. I suppose we can't revert because that would cause Linux rebuilds | 12:04:02 |
| emily | I take it a wayland rebuild is out of the question at this point? | 12:04:12 |
 Vladimír Čunát | So we override withWayland to false on darwin? | 12:05:01 |
| Vladimír Čunát | (Does wayland even make sense on darwin?) | 12:05:24 |
| emily | it's probably not a good thing if ~everything doing withWayland ? lib.meta.availableOn stdenv.hostPlatform wayland needs special-casing for Darwin because nothing expects Wayland on Darwin | 12:05:35 |
| Vladimír Čunát | * So we override withWayland to false on darwin, at least for now? | 12:05:38 |
| emily | shrug
This package is vital for macOS wayland compostors.
I am currently working on a wayland compositor for macOS, and this would help me reduce reliance on downstream custom derivations of libwayland.
X11 on macOS is a thing, Wayland on macOS could be too. in practice it is not
| 12:06:04 |
| Vladimír Čunát | In that case we'd... remove darwin from wayland.meta.platforms probably | 12:06:08 |
| emily | so I'm not inclined to put much work into it to keep Darwin working. | 12:06:11 |
| emily | right, this PR was specifically meant to make it build on Darwin again. | 12:06:22 |
| Vladimír Čunát | That's one thing. | 12:06:41 |
| emily | https://github.com/NixOS/nixpkgs/pull/492078 | 12:10:33 |
| emily | haven't tested the qtbase build, but I'm pretty sure this is the root cause | 12:10:42 |
| emily | looking at scikit-build-core | 12:11:22 |
| emily | the issue with scikit-build-core isn't the presence or lack of lipo; it can handle that | 12:12:18 |
| emily | it's https://pytest-subprocess.readthedocs.io/en/1.4.0/usage.html#unregistered-commands, "By default, when the fake_process fixture is being used, any attempt to run subprocess that has not been registered will raise the ProcessNotRegisteredError exception." | 12:12:41 |
| emily | hexa: any recent changes to pytest stuff? | 12:12:48 |
| emily | uh whoops, I included my random testing of scikit-build-core in the Wayland commit :) | 12:13:14 |
| emily | oh for goodness sakes | 12:17:00 |
| emily | https://github.com/scikit-build/scikit-build-core/blob/2ddfb9f436fcfceb404d6a3444f7d6ba1093ea60/tests/test_get_requires.py#L34-L37 | 12:17:04 |
| emily | their fixture just won't work if you built single-arch binaries like we do | 12:17:37 |
| emily | oh, I misunderstood what they were doing… | 12:24:28 |
 K900 | How did we merge that wayland PR wtf | 12:34:07 |
| Vladimír Čunát | meta.availableOn is easy to miss | 12:35:56 |
| emily | ok, we just need to backport the upstream fix. | 12:36:21 |
| emily | the janky downstream patch was already in-tree to be fair | 12:37:10 |
| emily | it just didn't actually build | 12:37:16 |
| emily | fwiw it seems bad that libX11 and wayland and etc. leaking into the qtbase build environment can affect features | 12:37:42 |
