| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
| |
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
| |
Also depend on >=egl-wayland-1.1.10 given libnvidia-vulkan-producer
now has wlEglCreateSurfaceExport
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
| |
Now needs >=egl-wayland-1.1.10 or else get a missing symbol if try to
use vulkan on wayland. 515.57 still ships with 1.1.9 but it's a
modified version of 1.1.9, likewise 1.1.10-r1 is a bit modified given
it's currently in a poor state upstream (hopefully this works fine).
Also install nvidia-dbus.conf again, file was sanitized upstream
and believe it should be fine install.
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
| |
Overlooked updating that one.
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
| |
Was waiting for a bump to do these revbumps given 1.1.10 was not
keyworded anyway. 1.1.10 needs wlEglCreateSurfaceExport from
libnvidia-vulkan-producer.so which currently only exist in >=515.57
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
| |
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
| |
Note that even though it is newer (for now), it is not recommended
to use the 0/vulkan branch unless you are involved in vulkan
application development, upstream bugs may not get fixed (aka it
adds vulkan features on top of old drivers and rebase occasionally).
Please report if packaging issues, as don't check masked 0/vulkan much.
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
|
| |
Was to help reduce surprises when profile removed it as default,
but been a while and should be little meaning to this now.
README.Gentoo still warns about this for new installs.
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Branch been tested for a while and major known issues are sorted
out in stable now.
May at most still be a bit rocky for wayland users given the
some applications will switch to using GBM, but gnome/kde have
rolled out fixes to major issues a while ago.
Owners of Kepler cards should mask >=x11-drivers/nvidia-drivers-471
as the ebuild will warn about (thanks for testing wrt #837317).
Closes: https://bugs.gentoo.org/837317
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
| |
Upcoming egl-wayland-1.1.10 is known incompatible, meant to bump
it early to fix wayland vulkan issues with 0/515 but it requires
yet unreleased nvidia-drivers that'd have the missing symbol in
the vulkan-producer lib.
And so .10 will likely never work with old 0/510, and there should
be no reason to use 1.1.7 with it (barely tested combination).
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
| |
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
| |
That this works without eval feels a bit uncanny because the
backslash for the following space is inside the variable, but
it evaluates whole thing together.
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
New feature in >=kernel-5.18, believe few have enabled it in Gentoo
but it's been causing issues in binary distros that tried to enable
it by default.
Seems possible to workaround, but would rather wait for nvidia's fix.
Add to non-5.18 ebuilds too for consistency and not losing track,
is quiet if the option doesn't exist.
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
| |
Unlike CXX, this did not error out but still fails to use objdump
with -native-symlinks. Only 515 is affected, but sync all to have
less ebuild differences and have one less thing to worry about.
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This isn't specific to nvidia-drivers (even a blank test.c
fails), but it's not uncommon this is the only third party
module that a user have and a warning wouldn't hurt.
Normally not an issue given gcc plugins are typically not
enabled by default (or the option is set, but individual
plugins are disabled) and mismatching is mostly fine.
Also note about kernel-bin given the module-specific bits
/can/ be rebuilt and it may sound confusing otherwise.
Bug: https://bugs.gentoo.org/849011
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
|
| |
Turns out this is only needed if users manually enabled
DynamicPowerManagement (not default) and feel this should not be
installed by default (no revbump given not known to be harmful
and will be cleaned on kernel upgrades).
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
| |
Drop CONFIG_BACKLIGHT_CLASS_DEVICE check, nvidia added its own
proper check for it allowing to disable again.
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
| |
Not something I experimented much with, but I've heard reports
it can work badly even with systemd services. It may or may not
work better, but avoid sounding like it's always recommended.
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
| |
Only 515 is affected, but sync all to have less ebuild
differences and have one less thing to worry about.
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
|
| |
As upstream does, believe this should be setup manually if needed and
the rules are questionable. Unlikely many are using powerd yet (many
distros don't ship it too), hardly hear of it working properly. So
skipping a warning and letting it propagate in kernel module rebuilds.
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
| |
Bug: https://bugs.gentoo.org/845063
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
| |
Bug: https://bugs.gentoo.org/845063
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
| |
Bug: https://bugs.gentoo.org/845063
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
| |
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
| |
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
| |
linux-mod already checks for `use driver`
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
User reported it solved some power management issues (e.g. oddities
followed by Xorg segfault when using sleep with GNOME), and other
distros like Fedora install these unconditionally. Only for 470+
given wasn't included in nvidia's README for 390.xx and likely not
relevant.
Skip revbump, not known(?) to affect many and will propagate on
kernel upgrades + rebuild.
Thanks-to: Randall T. Vasquez
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
| |
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
| |
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
| |
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
| |
Already mentioning it's experimental a lot, but may not be clear
that features are straight up missing.
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
| |
Closes: https://bugs.gentoo.org/843827
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
| |
Not everyone recognize these code names, and will end up with
broken drivers on the wrong card.
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
|
|
| |
- use full source rather than bundled kernel-open which comes
with a prebuilt .o_binary similarly to closed source blob
- enable NVreg_OpenRmEnableUnsupportedGpus=1 by default or
else it's unlikely to work on the average GPUs
- symlink application profile to help nvidia-settings start
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Opting to leave betas unkeyworded after all, feel free to test.
- new IUSE=kernel-open to build open source variant, entirely
untested (lacking a Turing / Ampere card). May or may not
mask the USE entirely given how experimental it is, but
given unkeyworded not worrying about this yet.
- CONFIG_BACKLIGHT_CLASS_DEVICE conftest was removed and is
used unconditionally, depend on it until further notice.
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
| |
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
| |
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
| |
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
| |
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Should revbump but would rather not cause rebuilds over this
given this is a no-op unless forced USE=-pie on gcc through
profile's package.use.force (clang now also defaults to it).
nvidia-drivers already get rebuilt on kernel upgrades which will
phase it out for -pie users (please manually rebuild if issues).
Legacy 0/390 branch is not affected.
Closes: https://bugs.gentoo.org/840389
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
| |
Switching to egetent wasn't very useful given the limited use-cases
where nvidia-drivers can be installed (and there's no egetgroupid)
but, given there's talk to improve its ROOT / crossdev support, it
may be good to finally change this.
Also add sanity-check to ensure we got a number, still skip assert
so failure will give the sanity-check's error message.
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
| |
"wait until NVIDIA releases a fixed version" may not be clear enough.
This further hold true for _rc/bleeding/.0/.1 kernels, would consider
for non-legacy drivers if didn't work with the current-stable LTS
kernel at most (generally LTS is recommended if using any out-of-tree
modules, or at least wait a bit before using new major kernels).
Bug: https://bugs.gentoo.org/838118
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
| |
510.60.02 been uneventful and gnome issues weren't related to it,
shouldn't need to keep this version.
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
| |
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
| |
Also EAPI=8 and sync NV_PIN with latest 470 (this is not actively
tested for masked 0/vulkan, please report if issues).
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Now that hwloc depends on this by "default" when VIDEO_CARDS=nvidia,
not being enabled by default tend to be an annoyance.
This only has a cheap build-only dependency on libX11/Xext, is tiny,
and doesn't take long to build. Only makes some sense to disable if
wayland-only or headless without X (albeit this is why the USE exists
and isn't unconditional, /could/ be merged with USE=X although that
USE has more implications.. may revisit later).
Bug: https://bugs.gentoo.org/836004
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
|
|
|
|
| |
It may have been better to be USE=nvctrl or merged with USE=X, but
let's at least provide a description so it's clearer what it's for.
Bug: https://bugs.gentoo.org/836004
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
| |
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|
|
|
|
| |
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
|