Jakub Konka

Hmm, I don't see the sysroot set though. It needs to be set for Zig to pass it on to the linker as `-syslibroot` which is then required to work...

No inconvenience at all. Yes, it is. You should set it to the same value to what you prepend in `xSdkPath` call.

> @kubkon I realize now we're facing this issue as well: > > * [flac: linux -> mac cross compilation broken after Zig upgrade hexops/mach#1009](https://github.com/hexops/mach/issues/1009) > * [opus: linux ->...

FYI my comments regarding `-flat_namespace` are completely wrong and unrelated - dylibs are allowed to re-export dylibs in both `-twolevel_namespace` and `-flat-namespace`. Boy, it just took me rewriting the linker...

@slimsag OK, confirmed locally on Linux with the following changes to `glfw` project: ```diff diff --git a/build.zig b/build.zig index e63356f3..a936d992 100644 --- a/build.zig +++ b/build.zig @@ -82,14 +82,6 @@ pub...

@andrewrk I believe https://github.com/ziglang/zig/pull/18207 should fix the CI for you.

Do we actually need this at all? Would it not be better to promote having `libc.txt` setup instead? Zig should parse `libc.txt` inferring sysroot and all relevant system include/lib dirs...

As a general direction, we do not want to expose `-fuse-ld` or `-B` flags to the user. Instead we would appreciate patches and issues submitted on GH.

Also please note that `-fno-lld` and `-flld` are only temporary - once we support more target combinations we will always link using our linkers.

Firstly, if we cross-compile, Zig doesn't autodetect sysroot and system include paths. It is up to the user to provide all of those via `build.zig` or otherwise. However, this has...