Arch Linux User Repository

For issues where meson says that nvcc can't compile programs:

Can you source /etc/profile.d/cuda.sh and try again? I had this issue as well and in the end it seemed to have been about environment variables

I think you need to modify line 22 to prevent the following error.

arch-meson "$pkgname-$pkgver" builddir -Db_pie=false -Dcuda_link_args=[] -Db_lto=false

Error:

ninja: Entering directory `/build/wicuvanity/src/builddir'
[1/2] Compiling Cuda object wicuvanity.p/main.cu.o
[2/2] Linking target wicuvanity
FAILED: [code=1] wicuvanity 
nvcc  -o wicuvanity wicuvanity.p/main.cu.o -Wl,-O1 -Wl,--sort-common -Wl,--as-needed -Wl,-z,relro -Wl,-z,now -Wl,-z,pack-relative-relocs -flto=auto
nvcc fatal   : Unknown option '-Wl,-O1'
ninja: build stopped: subcommand failed.
INFO: autodetecting backend as ninja
INFO: calculating backend command to run: /usr/bin/ninja -C /build/wicuvanity/src/builddir

Build fails for me:

Found ninja-1.13.2 at /usr/bin/ninja
ninja: Entering directory `/build/wicuvanity/src/builddir'
[1/2] Compiling Cuda object wicuvanity.p/main.cu.o
[2/2] Linking target wicuvanity
FAILED: [code=1] wicuvanity 
nvcc  -o wicuvanity wicuvanity.p/main.cu.o -Wl,-O1 -Wl,--sort-common -Wl,--as-needed -Wl,-z,relro -Wl,-z,now -Wl,-z,pack-relative-relocs -flto=auto
nvcc fatal   : Unknown option '-Wl,-O1'
ninja: build stopped: subcommand failed.
INFO: autodetecting backend as ninja
INFO: calculating backend command to run: /usr/bin/ninja -C /build/wicuvanity/src/builddir

Good catch, that's because meson falls back to cmake for finding CUDAToolkit. I've added cmake to the makedepends.

I was having troubles building in a clean chroot environment (with aurutils), I got the following error message:

Found CMake: NO
Run-time dependency cudatoolkit found: NO (tried pkgconfig)

meson.build:8:8: ERROR: Dependency "CUDAToolkit" not found, tried pkgconfig

I was able to fix it by adding cmake to the makedepends. Maybe you could include it in the next release ?

Hmm yeah probably, but that should be the job of your session manager to source...

I saw your comment after my last comment: sourcing /etc/profile.d/cuda.sh does indeed work. Can't that be done from within the PKGBUILD maybe?

I was eventually able to build it in an arch docker container after installing cuda and gcc-13, removing gcc then symlinking gcc-13/g++-13 to gcc/g++, and finally commenting out the CUDAToolkit dependency in the meson.build.

CUDA's got to be the neediest framework for sure!

For issues where meson says that nvcc can't compile programs:

Can you source /etc/profile.d/cuda.sh and try again? I had this issue as well and in the end it seemed to have been about environment variables

Hi! This doesn't seem to be building on Arch:

==> Making package: wicuvanity 1.0-1 (Fri 01 Nov 2024 23:44:01)
==> Checking runtime dependencies...
==> Checking buildtime dependencies...
==> Retrieving sources...
  -> Found wicuvanity-1.0.tar.gz
  -> Found wicuvanity-1.0.tar.gz.asc
==> Validating source files with sha256sums...
    wicuvanity-1.0.tar.gz ... Passed
    wicuvanity-1.0.tar.gz.asc ... Skipped
==> Verifying source file signatures with gpg...
    wicuvanity-1.0.tar.gz ... Passed
==> Extracting sources...
  -> Extracting wicuvanity-1.0.tar.gz with bsdtar
==> Removing existing $pkgdir/ directory...
==> Starting build()...
The Meson build system
Version: 1.6.0
Source dir: /tmp/wicuvanity/src/wicuvanity-1.0
Build dir: /tmp/wicuvanity/src/wicuvanity-1.0/builddir
Build type: native build
Project name: wicuvanity
Project version: 1.0

meson.build:1:0: ERROR: Compiler nvcc cannot compile programs.

A full log can be found at /tmp/wicuvanity/src/wicuvanity-1.0/builddir/meson-logs/meson-log.txt
==> ERROR: A failure occurred in build().
    Aborting...