| 3 Sep 2023 |
 Artturin | i think.. | 20:44:24 |
 p14 | For git and curl it's an odd one. It does a link check which fails because of the missing transistive dependencies. If you fix that, it subsequently runs curl-config which reports the correct set of things to link against :/ | 20:44:36 |
| p14 | i.e. it seems like git's config test is just broken. But it's not the only one, there are plenty of other packages broken in that they don't know what to link against. | 20:45:15 |
|  problems joined the room. | 21:33:29 |
| 4 Sep 2023 |
|  Robert Hensing (roberth) changed their display name from Robert Hensing (roberth) to roberth. | 15:27:30 |
| Robert Hensing (roberth) changed their display name from roberth to Robert Hensing (roberth). | 15:28:17 |
| problems | oh! this room exists! | 19:41:30 |
| problems | i was talking in #exotic:nixos.org about wanting to explore "wasi-native" nixpkgs, where every build is cross compiled from wasm64-wasi to your native architecture. that way, it wouldn't matter what your host is or if you're compiling "natively" or "cross" and you would be able to use heterogenous architectures to do distributed builds | 19:44:39 |
| problems | i think the first step of this is creating a stdenv for wasi, but i have no idea where to start with that | 19:44:58 |
| problems | * i was talking in #exotic:nixos.org about wanting to explore "wasi-native" nixpkgs, where every build is cross compiled from wasm64-wasi to your native architecture (using boot.binfmt.emulatedSystems = [ "wasm64-wasi" ]). that way, it wouldn't matter what your host is or if you're compiling "natively" or "cross" and you would be able to use heterogenous architectures to do distributed builds | 19:45:47 |
| 5 Sep 2023 |
 trofi | For linux there pkgs/stdenv/linux/ and make-bootstrap-tools.nix which builds a set of minimal binaries to start from. | 02:40:49 |
| trofi | Alternatively you can coerce pkgs/stdenv/default.nix to boot from system's binaries (impure or native build): might want to float stagesNative above. | 02:42:41 |
|  @andreas.schraegle:helsinki-systems.de left the room. | 15:00:42 |
| 11 Sep 2023 |
|  Arthur joined the room. | 02:13:22 |
| 13 Sep 2023 |
 Winter (she/her) | Hi all, poking my head in to confirm some weird thing. Is it expected that libclang, if added to a derivation's nativeBuildInputs (strictDeps = false at that), will more or less completely blow up the ability to use Clang? (e.g. not being able to find any libc headers like stdbool.h.) | 22:07:11 |
| Winter (she/her) | I can only repro with a specific CMake project, trying to figure out a way to reduce the issue. | 22:10:02 |
| Winter (she/her) | The reason I ask here is because it doesn't happen with buildInputs. | 22:10:29 |
| Winter (she/her) | The first suspect of CMAKE_INCLUDE_PATH differences isn't the case, hm. | 22:25:00 |
| Artturin | In reply to @winterqt:nixos.dev
Hi all, poking my head in to confirm some weird thing. Is it expected that libclang, if added to a derivation's nativeBuildInputs (strictDeps = false at that), will more or less completely blow up the ability to use Clang? (e.g. not being able to find any libc headers like stdbool.h.) Diff the dumped env vars | 22:31:35 |
| Artturin | * In reply to @winterqt:nixos.dev
Hi all, poking my head in to confirm some weird thing.
Is it expected that
libclang, if added to a derivation's nativeBuildInputs (strictDeps = false at that), will more or less completely blow up the ability to use Clang? (e.g. not being able to find any libc headers like stdbool.h.)
Diff the dumped env vars after the setup | 22:31:55 |
| Winter (she/her) | In reply to@artturin:matrix.org
Diff the dumped env vars That's what I'm doing, and nothing is different (after sorting + de-duping) that I've seen. | 22:31:56 |
| Winter (she/her) | It's definitely CMake-specific, but I'm still trying to figure out how, none of the CMAKE_* env vars are different in a substantial way (unless order matters, which it shouldn't). | 22:32:39 |
| Artturin | Is this just the normal package set and not a pkgsLLVM or so | 22:33:00 |
| Winter (she/her) | It's pkgsLLVM bc Darwin :/ | 22:34:58 |
| Winter (she/her) | FWIW I don't need it to be in nBI, I'm just trying to figure out why it breaks. | 22:35:23 |
| Winter (she/her) | (specifically because mkShell.packages won't work then) | 22:35:36 |
| Artturin | So build != host? | 22:35:41 |
| Artturin | In reply to @winterqt:nixos.dev
(specifically because mkShell.packages won't work then) .packages is bad anyways and the correct dependency attributes should be used | 22:36:50 |
| Winter (she/her) | In reply to @artturin:matrix.org
So build != host? build == host, this is a darwin build in a darwin machine + no strictDeps | 22:50:20 |
| Artturin | So it's not pkgsLLVM | 22:51:52 |
