| 16 Jun 2025 |
K900 | There's LAYERS | 05:33:42 |
emily | well it's about finding binaries right? | 05:33:46 |
emily | or no? | 05:33:50 |
K900 | Not quite | 05:34:15 |
K900 | It's not really finding specific things | 05:34:29 |
K900 | It's finding "Qt" | 05:34:32 |
K900 | Whatever that means | 05:34:37 |
K900 | If you want to dig into it, give me another like hour | 05:37:38 |
K900 | So I can add all the nativeBuildInputs everywhere | 05:37:50 |
K900 | And then I can push it and you can see what happens if you take those out | 05:37:58 |
K900 | OK I no longer understand anything | 08:30:43 |
K900 | Some Qt packages are fine being in buildInputs | 08:30:50 |
K900 | Some only get found from nativeBuildInputs | 08:30:58 |
K900 | e.g. qtmultimedia wants to be in nativeBuildInputs, but qtsvg is fine being in buildInputs | 08:31:21 |
K900 | Neither contains a single executable | 08:31:27 |
K900 | angelfish only builds when qtwebengine is in both | 08:34:41 |
emily | anything obviously different about the CMake logic searching for them? | 09:08:32 |
K900 | No idea | 09:24:20 |
K900 | I have not looked yet | 09:24:22 |
K900 | I am still bashing my head against getting it all to even build again | 09:24:30 |
K900 | Down to only a few packages though | 09:24:46 |
K900 | @emily https://github.com/NixOS/nixpkgs/pull/416534/commits/a9b16de0665b88d0d428e513a33144bdb865732c | 09:39:58 |
K900 | Basically if you take this and just revert everything but the ECM change, you'll see it blow up | 09:40:13 |
emily | I'd look into it if I had the time :( but that commit makes it look like it's just qttools? is there something else other than QDoc going on? | 09:47:35 |
K900 | Adding qttools to nativeBuildInputs makes the FindQt machinery explode | 09:51:31 |
K900 | Somehow | 09:51:39 |
K900 | I don't fully understand how or why | 09:51:43 |
emily | but propagating it as a native input does not? | 09:52:07 |
K900 | Propagating it also does | 09:52:14 |
K900 | Like it works with no qttools | 09:52:25 |