Arch Linux User Repository

@failsafe65, thanks for the patch. It's a little hard to follow the changes with all the whitespace updates, but it does like the more generalized fix was more complicated than just updating the PKGBUILD. Everything's working here on my end, finally updating to GCC 15, with a 3050 Mobile in discrete graphics mode, 12 processes in nvidia-smi, and laptop suspend still working as normal :)

Is there a plan to incorporate these patch steps into the package as a whole? I'm trying to decide if I should do them or wait for the package to be updated.

@ww998: I got some idea and build succeed with gcc version 15.1.1 20250425 (GCC)

1. download 0002-CFLAGS-Set-std-gnu17-for-all-compilation-flags.patch file and add into PKGBUILD
2. add patch command in prepare function
3. just make it

Here is diff for PKGBUILD file

diff --git a/PKGBUILD b/PKGBUILD
index 75f4214..9a4cd94 100644
--- a/PKGBUILD
+++ b/PKGBUILD
@@ -20,12 +20,14 @@ source=('nvidia-drm-outputclass.conf'
         'nvidia.rules'
         'systemd-homed-override.conf'
         'systemd-suspend-override.conf'
+        '0002-CFLAGS-Set-std-gnu17-for-all-compilation-flags.patch'
         "https://us.download.nvidia.com/XFree86/Linux-x86_64/${pkgver}/${_pkg}.run")
 sha512sums=('de7116c09f282a27920a1382df84aa86f559e537664bb30689605177ce37dc5067748acf9afd66a3269a6e323461356592fdfc624c86523bf105ff8fe47d3770'
             '4b3ad73f5076ba90fe0b3a2e712ac9cde76f469cd8070280f960c3ce7dc502d1927f525ae18d008075c8f08ea432f7be0a6c3a7a6b49c361126dcf42f97ec499'
             'f8f071f5a46c1a5ce5188e104b017808d752e61c0c20de1466feb5d693c0b55a5586314411e78cc2ab9c0e16e2c67afdd358da94c0c75df1f8233f54c280762c'
             'a0183adce78e40853edf7e6b73867e7a8ea5dabac8e8164e42781f64d5232fbe869f850ab0697c3718ebced5cde760d0e807c05da50a982071dfe1157c31d6b8'
             '55def6319f6abb1a4ccd28a89cd60f1933d155c10ba775b8dfa60a2dc5696b4b472c14b252dc0891f956e70264be87c3d5d4271e929a4fc4b1a68a6902814cee'
+            '6814990f8046759d35f724ac9114d7fa284710fd1ad8cca7e1a861ea54a72fe4f67b5614f157a911f5ecfa0c964fa989edc61c85b1cfef6428e0cd7cdeea856e'
             '676b1de35d21e80091528a49080c114e0870ea90b3f3721265ae8071abbc2183c851e6f11ba96a332c743fedfaf4ec9a014ad6ceed586fdbb03d94d33405e356')


@@ -40,6 +42,7 @@ create_links() {
 }

 prepare() {
+    local buildroot=$PWD
     sh "${_pkg}.run" --extract-only
     cd "${_pkg}"
     bsdtar -xf nvidia-persistenced-init.tar.bz2
@@ -62,6 +65,9 @@ DEST_MODULE_LOCATION[4]="/kernel/drivers/video"' dkms.conf

     # Gift for linux-rt guys
     sed -i 's/NV_EXCLUDE_BUILD_MODULES/IGNORE_PREEMPT_RT_PRESENCE=1 NV_EXCLUDE_BUILD_MODULES/' dkms.conf
+
+    # patch for gcc 15
+    patch -i ${buildroot}/0002-CFLAGS-Set-std-gnu17-for-all-compilation-flags.patch
 }

 package_opencl-nvidia-550xx() {

@jadespace: Thank you very much, I now managed to install the 550 driver. I guess there must be a much more elegant way to do it, but maybe this can help other people who are lost here:

download the packagebuild

yay -G nvidia-550xx-dkms

install from the packagebuild. this will fail but download all needed files

cd nvidia-550xx-dkms; makepkg -i ./PKGBUILD

unpack the downloaded nvidia driver package

download the patch from @Fuxino

search for Kbuild and conftest.sh in the extracted files

Apply the patch with the "patch" command

Repack the package, keep the original filename

navigate back to the dir with the PKGBUILD

Use makepkg again to now succesfully install the patched package

makepkg -i ./PKGBUILD

I added the nvidia-550xx-* packages to the IgnorePkg line in /etc/pacman.conf to prevent further updates as suggested by @jadespace I did not have to patch other packages. Everything seems to work now.

I sadly have to use the 550 branch because every driver after this one has a showstopper bug for me that makes it impossible to use. As this affects my notebook there is no way to switch to another GPU.

I hope this AUR package will continue to find some love.

@ww998, it would be applied similarly to the commit that contained the patch. That is, for me having installed it with paru, the build flags are being patched in kernel and kernel-open under $HOME/.cache/paru/clone/nvidia-550xx-dkms/src/NVIDIA-Linux-x86_64-550.163.01.

The slightly painful part is that installing it would likely mean needing to update the PKGBUILDs for all the other 550 driver packages that you have installed (for example, I have to uninstall and reinstall Steam every time there's an update here because of circular dependencies).

Judge whether it's worth your time to pursue those manual patches and any subsequent build errors that arise from changes in GCC 15, or wait for the packages to get updated. (@failsafe65, do you need any co-maintainers?)

In the meantime, I downgraded back to GCC 14 with everything working as normal:

sudo pacman -U /var/cache/pacman/pkg/gcc-14.2.1+r753+g1cd744a6828f-1-x86_64.pkg.tar.zst /var/cache/pacman/pkg/gcc-libs-14.2.1+r753+g1cd744a6828f-1-x86_64.pkg.tar.zst

An additional tip for anyone that needs it: if you add gcc gcc-libs to the IgnorePkg line in /etc/pacman.conf, pacman will skip over those two until you're ready for them to be updated.

@Fuxino : could you please explain how to apply the patch?. I don't understand what to to with it. It fails when I apply it to the PKGBUILD of nvidia-550xx-dkms. thank you very much.

The module fails to build with GCC 15, this patch is needed: https://gitlab.archlinux.org/archlinux/packaging/packages/nvidia-utils/-/blob/11deaedbd194cacd732c44a902699a56d2d59d4b/0002-CFLAGS-Set-std-gnu17-for-all-compilation-flags.patch

Bug fixed as of 550.163.01

There is a new version: 550.163.01 I can't test it this weekend, maybe they fixed this latest bug. It is not on the changelog though.

With the update of linux to 6.14.2 in main, this issue is hitting more people.

Putting

sed -i '/\.date/d' nvidia-drm/nvidia-drm-drv.c

just after cd kernel in PKGBUILD fixed the issue for me.

Just for future reference, running the 550 series fixes two important bugs for me:

kwin_wayland memory leak: https://bugs.kde.org/show_bug.cgi?id=496469#add_comment

External Monitor not waking up after suspend if refresh rate is higher than 30Hz: https://forums.developer.nvidia.com/t/bug-external-monitor-fails-to-wake-up-from-powersave-mode-if-refresh-rate-is-higher-than-30hz/316612/21