Sender | Message | Time |
---|---|---|
25 May 2025 | ||
it uses libcxx for WASM, I guess? but does it use anything else from pkgsCross ? | 02:50:52 | |
Yeah, idk | 02:51:04 | |
I think it needs wasilibc | 02:52:01 | |
Ok yeah, it needs a wasm toolchain. It looks like upstream uses the wasi-sdk. | 02:54:37 | |
How does crossStdenv handle cross-compiled dependencies? DXVK needs SPIRV and Vulkan headers and windows.pthreads . | 13:39:07 | |
Also, how does it handle NIX_ vars? Does this even work with Darwin since Darwin needs an SDK? | 13:44:42 | |
In reply to @reckenrode:matrix.orgSuffix salt should work with the env vars. Idk about Darwin but it could be made to work. | 14:28:05 | |
In reply to @reckenrode:matrix.orgYou'll have to override the stdenv for the dependencies. | 14:28:23 | |
In reply to @rosscomputerguy:matrix.orgThat doesn’t seem like an improvement. | 14:29:31 | |
In reply to @reckenrode:matrix.orgI don't see really any other way of doing cross without either creating a new nixpkgs instance (what we already have) or only provide cross compilation stdenv's | 14:30:30 | |
DXVK uses a custom MinGW (with the Win32 thread model instead of the POSIX one). When you build https://github.com/NixOS/nixpkgs/blob/6fb6fd3ceff99de72d881f1a97773647c94e5a82/pkgs/by-name/dx/dxvk/package.nix#L28-L44 DXVK itself has dependencies: https://github.com/NixOS/nixpkgs/blob/6fb6fd3ceff99de72d881f1a97773647c94e5a82/pkgs/by-name/dx/dxvk_2/package.nix#L74-L81 The replacement sounds like I need to manually enumerate the transitive dependencies of DXVK, replace all of their | 15:08:53 | |
crossStdenv is still expensive, but now the maintainability of my package has gone down because I have more work to do to build my package instead of just referencing pkgsCross.mingw32.dxvk_2 and pkgsCross.mingwW64.dxvk_2 . | 15:09:59 | |
Note that it’s not possible to write a function to replace inputs that works in all cases because not all dependencies are in buildInputs (or propagatedBuildInputs ). It’s the same problem that overrideSDK had. | 15:15:22 | |
* Note that it’s not possible to write a function to replace inputs that works in all cases because not all dependencies are in buildInputs (or propagatedBuildInputs ). It’s the same problem that overrideSDK had. You’ll always need to know what dependencies your inputs have to make sure there are no problems. | 15:15:44 | |
In reply to @reckenrode:matrix.orgYeah, I have thought about a recursiveStdenvOverride function which might work here | 15:30:31 | |
It would go through and apply the new stdenv where it should apply. | 15:30:54 | |
That can’t be done in the general case. Would it be possible to have a function like yesReallyDoInstantiateAnotherNixpkgsWithStdenc for non-trivial cases? | 15:31:47 | |
* That can’t be done in the general case. Would it be possible to have a function like yesReallyDoInstantiateAnotherNixpkgsWithStdenv for non-trivial cases? | 15:31:59 | |
Idk | 15:32:48 | |
Thinking about Wine, it has a number of vendored dependencies that would be nice to devendor. It would also be nice to build wine-mono and wine-gecko from source. | 15:32:54 | |
I would not want to have to manually override all the dependencies to build Gecko or Mono. | 15:33:22 | |
Yeah, it seems like no matter what, cross compilation is always going to be expensive but necessary. | 15:34:39 | |
Ericson brought up that crossStdenv could be done easier with the GCC rewrite he's been trying to push | 15:37:23 | |
And there's things crossStdenv uses that should have us make changes to the CC wrapper and the toolchain attributes PR would make things nicer as well. | 15:38:13 | |
It makes sense to have a crossStdenv for trivial cases that just need a compiler, but sometimes a package set is actually needed. A stdenv adapter won’t be good enough as a substitute. Darwin went through that with the SDK. Overriding the SDK in Rust packages was a major problem and often didn’t work. | 15:41:27 | |
Also note that such an adapter can’t be written recursively. My first attempt at a general adapter used recursion. Eval performance increased by over 500%. I had to use builtins.genericClosure to do it with good performance. | 15:42:07 | |
Ideally, it would be nice if we could have a way to "pull back" the stdenv to stage 3 and then inject new stages in. | 15:43:58 | |
Then we could overlay that and it hopefully wouldn't have to eval stages before the new stages. | 15:44:26 | |
I’d prefer to view the stdenv bootstrap as a black box so that implementation details don’t become fixed API. | 15:47:48 | |
True though we'd have to do something to the stdenv for it to be able to do cross unless we split CC away from the stdenv. | 15:48:47 |