Arch Linux User Repository

I reported my finding here: https://forums.developer.nvidia.com/t/nvidia-modules-470-239-06-build-failure-with-gcc-14-due-to-conftest-sh/292645

@benjarobin: can confirm, your patches works.

I improved the second patch to properly fix warning during conftest related to EFI (instead of hiding it).

To build a package with a proper fix, do the following:

git clone https://aur.archlinux.org/nvidia-470xx-utils.git
cd nvidia-470xx-utils
curl https://0x0.st/XKB6.patch -o 0001-Fix-GCC-14-build.patch
patch -p1 -i 0001-Fix-GCC-14-build.patch
makepkg

See fixes provided by: https://0x0.st/XKB6.patch

Using -fpermissive is wrong, and introduces a major bug: The call to vmap is done with the wrong function call (missing one argument), as explained in [2].

I did create various patches in order to fix it properly, see [1] and the Nvidia fix that was back-ported [3].

To build a package with a proper fix, do the following:

git clone https://aur.archlinux.org/nvidia-470xx-utils.git
cd nvidia-470xx-utils
curl https://0x0.st/XKmG.patch -o 0001-Fix-GCC-14-build.patch
patch -p1 -i 0001-Fix-GCC-14-build.patch
makepkg

[1] https://0x0.st/XKmG.patch

[2] https://forums.developer.nvidia.com/t/nvidia-modules-build-failure-with-upcoming-gcc-14-and-recent-kernels-due-to-misfiring-conftest-sh-test-heads-up/279072

[3] https://github.com/NVIDIA/open-gpu-kernel-modules/commit/811073c51e1f098834e90e67a074f1746b6e2831#diff-32da369dcd8cd223fa47c8b86028f1a7572ed699c33f2057f4702e4b36f13dd5

@a821, @Mechanicus: actually i tried easiest workaround which is downgrade ['gcc-libs', 'gcc'] from 14.1.1+r1+g43b730b9134-1 to 13.2.1-6 and has been compiled and worked again as previously. Thanks you folks for your efforts to find a workarounds.

loading packages...
warning: downgrading package gcc (14.1.1+r1+g43b730b9134-1 => 13.2.1-6)
warning: downgrading package gcc-libs (14.1.1+r1+g43b730b9134-1 => 13.2.1-6)
resolving dependencies...
looking for conflicting packages...

Package (2)  Old Version               New Version  Net Change

gcc          14.1.1+r1+g43b730b9134-1  13.2.1-6     -17.55 MiB
gcc-libs     14.1.1+r1+g43b730b9134-1  13.2.1-6      -3.81 MiB

==> dkms install --no-depmod nvidia/470.239.06 -k 6.6.30-2-lts
==> depmod 6.6.30-2-lts
(6/6) Updating linux initcpios...
==> Building image from preset: /etc/mkinitcpio.d/linux-lts.preset: 'default'
==> Using default configuration file: '/etc/mkinitcpio.conf'
  -> -k /boot/vmlinuz-linux-lts -g /boot/initramfs-linux-lts.img
==> Starting build: '6.6.30-2-lts'
  -> Running build hook: [base]
  -> Running build hook: [udev]
  -> Running build hook: [microcode]
  -> Running build hook: [keyboard]

@unixv: Try -fpermissive patch. Copy the context to a file, for example workaround.patch, and then apply to the repo: git apply workaround.patch

diff --git a/PKGBUILD b/PKGBUILD
index 15fd45f..639ea2c 100644
--- a/PKGBUILD
+++ b/PKGBUILD
@@ -22,7 +22,8 @@ source=('nvidia-drm-outputclass.conf'
         "kernel-6.4.patch"
         "kernel-6.5.patch"
         "kernel-6.6.patch"
-        "kernel-6.1.76_6.6.15_6.7.3.patch")
+        "kernel-6.1.76_6.6.15_6.7.3.patch"
+        "gcc-permissive.patch")

 sha512sums=('de7116c09f282a27920a1382df84aa86f559e537664bb30689605177ce37dc5067748acf9afd66a3269a6e323461356592fdfc624c86523bf105ff8fe47d3770'  #nvidia-drm-outputclass
             '4b3ad73f5076ba90fe0b3a2e712ac9cde76f469cd8070280f960c3ce7dc502d1927f525ae18d008075c8f08ea432f7be0a6c3a7a6b49c361126dcf42f97ec499'  #nvidia-470xx.sysusers
@@ -31,7 +32,8 @@ sha512sums=('de7116c09f282a27920a1382df84aa86f559e537664bb30689605177ce37dc50677
             'd9df8b13d5fbe4f456a31de3679fd11aca7cd88771f8f11a5cc8ab17bab05861823b26d2d467593e5b90967a2902db691ca832f09fe21a5975eb3e4d6275e00c'  #kernel-6.4 patch
             '40ea983c81851b8a20629a943f9692cc0e007c815f46dd3b63cf1d7a44ccbed1ac5f9a3110720de54b017b9f9c7f5cc534ec6e097bc02fa5bd1de6b0a730c803'  #kernel-6.5 patch
             'fa9985b0dd9d7a973019da88a40d7830ea53df83af2d71d498b6b9dde04c0c797570991239dafc30cc2ccabbb8e8a7c3fa1bf89f26d8e8a26e624d7a17e5a84b'  #kernel-6.6 patch
-            'e0756be173930faf6b5104562d8092ccf8967fba7bf1a97d8e0e86e76a5e0e944b41f24aaee98eb8c7b3ad995353f6acb20a0f6173b82ee718664293a6493fea') #kernel-6.1 patch
+            'e0756be173930faf6b5104562d8092ccf8967fba7bf1a97d8e0e86e76a5e0e944b41f24aaee98eb8c7b3ad995353f6acb20a0f6173b82ee718664293a6493fea'  #kernel-6.1 patch
+            '5338c833081300a1857a2ed3e89a9ba67397b68a65a4cb029f69001519fc18d2b0f4af0aad389a2a0c0ecc7fd513cc47030a2f3c53b5b185fad25053b6a1f59c') #gcc-permissive patch

 create_links() {
     # create soname links
@@ -49,6 +51,9 @@ prepare() {
     cd "${_pkg}"
     bsdtar -xf nvidia-persistenced-init.tar.bz2

+    # GCC-14 workaround
+    patch -p1 -i "$srcdir/gcc-permissive.patch"
+
     cd kernel

 #    patch -p1 -i "$srcdir/kernel-6.4.patch"
diff --git a/gcc-permissive.patch b/gcc-permissive.patch
new file mode 100644
index 0000000..7f20dbd
--- /dev/null
+++ b/gcc-permissive.patch
@@ -0,0 +1,11 @@
+diff -Naru a/kernel/Kbuild b/kernel/Kbuild
+--- a/kernel/Kbuild    2024-02-03 08:25:55.000000000 +0200
++++ b/kernel/Kbuild    2024-05-12 15:37:12.620937599 +0300
+@@ -73,6 +73,7 @@
+ EXTRA_CFLAGS += $(call cc-option,-Werror=undef,)
+ EXTRA_CFLAGS += -DNV_SPECTRE_V2=$(NV_SPECTRE_V2)
+ EXTRA_CFLAGS += -DNV_KERNEL_INTERFACE_LAYER
++EXTRA_CFLAGS += -fpermissive
+ 
+ #
+ # Detect SGI UV systems and apply system-specific optimizations.

@unixv: did you read any of the latest comments / workarounds? If so, what did you try?

it's failed to compile and I'm no longer able to use my nvidia-470xx card .

(5/6) Install DKMS modules ==> dkms install --no-depmod nvidia/470.239.06 -k 6.6.30-2-lts Error! Bad return status for module build on kernel: 6.6.30-2-lts (x86_64) Consult /var/lib/dkms/nvidia/470.239.06/build/make.log for more information. ==> WARNING:dkms install --no-depmod nvidia/470.239.06 -k 6.6.30-2-lts' exited 10 `

@SoftExpert, older GCC will also work with -fpermissive flag, it is supported since GCC-4 version