Nix Cross Compiling | 577 Members | |
| 129 Servers |
| Sender | Message | Time |
|---|---|---|
| 10 Mar 2025 | ||
| I'm currently working on moving some projects at work to build using Nix, things are working find for macOS and linux, but we also need to produce artefacts for windows. Specifically, we're bundling a python application using pyinstaller, and the mere involvement of python seems to cause a lot of issues as a lot of what it takes to build a python package involves running code on the buildPlatform which is intended for the hostPlatform. Is there any hope in being able to achieve this? I've managed to build C and C++ things for windows using pkgsCross.mingwW64 (though I'd also love to know if I should target one of the crt platforms instead), but python seems particularly problematic. Maybe some kind of windows based franken-builder that's able to natively run windows binaries under WSL or something? | 23:41:32 | |
| 11 Mar 2025 | ||
| Nix can execute Windows binaries within WSL | 00:18:56 | |
| alternatively, Wine can execute within a Nix build, I've been doing that recently | 00:20:37 | |
| and finally I'm working on native Nix on Windows, but we're many months away from actually building the first derivation | 00:21:49 | |
| @puffnfreshG'day g'day - yeah I need to look more into using wine for things. I have a good grasp how a linux VM operates but wine feels like a different beast | 00:44:38 | |
| 02:38:39 | |
| bisected down to https://github.com/NixOS/nixpkgs/issues/191152 confirmed I can get libffi and gmp building again by adding a && !targetPlatform.isAndroidhowver I don't understand why it matters? I'm building from nixos, there's nothing on /usr/include | 13:41:24 | |
| er, rather, bisected down to https://github.com/NixOS/nixpkgs/pull/356162 which fixes that issue | 13:42:12 | |
I guess there's some sysroot shenanigans I don't understand going on and having -nostdlibinc is messing with it? | 13:43:50 | |
hmm though... merely adding && !targetPlatform.isAndroid seems risky for non-nixos linux as /usr/include can leak there? | 13:49:30 | |
| Redacted or Malformed Event | 13:52:46 | |
| 13 Mar 2025 | ||
any reason why lib.systems.examples, examples is there | 05:51:13 | |
*
any reason why | 05:51:19 | |
*
any reason why | 05:51:37 | |
In reply to @phanirithvij:matrix.org It's easier to refer to existing example platforms than to write all the attrs, some platforms have more than just
| 05:55:54 | |
| 14:58:36 | ||
| Would it be possible to add a new default/example pkgsCross system that is basically the same as arm-embedded but with newlib-nano? My goal would be to get that GCC cached so that I could use a newer GCC since it seems that latest gcc-arm-embedded update (14) PR was denied. | 18:55:20 | |
| I don't know what would be involved beyond adding something to the examples file, but I would be willing to do it | 19:01:07 | |
| 15 Mar 2025 | ||
| 17:07:38 | ||
| I ran into https://github.com/NixOS/nixpkgs/issues/231537 while cross compiling from a x86 host to a rv64 target. I know it's fixed in unstable, but this makes me wonder about a more general problem: Shellcheck doesn't care what host arch it runs on. Theoretically, we could shellcheck the rv64 shell scripts with a host (x86) shellcheck, eliminating the need for a rv64 ghc. Is that even possible in nixpkgs? | 17:10:09 | |
| or am i just missing something obvious | 17:10:14 | |
| Hi, any haskell.nix experts here? Could you please share a minimal flake.nix that can compile my Haskell library to Android aarch64? Documentation did not help at all. | 17:10:48 | |
| nevermind, it is possible, i was just setting hostplatform/buildplatform wrong in colmena | 18:06:58 | |
| 16 Mar 2025 | ||
| 15:30:01 | ||
| 17 Mar 2025 | ||
| is qt cross just totally busted? I'm failing both aarch64-linux and x86_64-freebsd cross with errors related to multiple qtbases ending up in inputs | 03:22:00 | |
| I found https://github.com/NixOS/nixpkgs/pull/267311 but I don't parse how it would help in this case | 03:22:47 | |
| * is qt cross just totally busted? I'm failing both aarch64-linux and x86_64-freebsd cross for qt5.qtsvg with errors related to multiple qtbases ending up in inputs | 03:23:30 | |
| commenting out the propagatedBuildInputs/depsTargetTargetPropagated assignment on qmake-hook and that seems to fix some things but the ramifications... unknown lol. will run nixpkgs-review soon | 03:51:11 | |
In reply to @rhelmot:matrix.orgYup and there's a few pr which may fix it | 03:59:02 | |
Use the cross label and in:title qt5 | 03:59:24 | |