hwmixvolume requires gtk, as per both the README[1] and attempting to
run it on a system without the required dependencies:
$ hwmixvolume
Traceback (most recent call last):
File "/usr/bin/hwmixvolume", line 21, in <module>
gi.require_version('Gtk', '3.0')
~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^
File "/usr/lib/python3.13/site-packages/gi/__init__.py", line 125, in require_version
raise ValueError('Namespace %s not available for version %s' %
(namespace, version))
ValueError: Namespace Gtk not available for version 3.0
Hence, as it is ineffective without gtk, gatekeep its installation
behind the already existing gtk USE flag.
[1] https://github.com/alsa-project/alsa-tools/blob/d444519df40d52de3903513a14085a0619f688db/hwmixvolume/README
Signed-off-by: Rahul Sandhu <nvraxn@posteo.uk>
Merges: https://codeberg.org/gentoo/gentoo/pulls/27
Signed-off-by: Eli Schwartz <eschwartz@gentoo.org>
fbdev=1 will be needed if we ever want to support DRM_SIMPLEDRM in the
future for dist kernels like Fedora does but that we explicitly disable
in the kernel config we use from there. Albeit older or unsupported
drivers will continue to either lack support or keep it off-by-default
due to having more issues.
390.x + 470.x that lack fbdev=1 support will remain in the tree likely
until December 2027 after the last kernel to support these versions
will reach EOL given there is no plans to patch old drivers for newer
kernels in Gentoo at the moment. 535.x + 570.x will likely be removed
before then whenever NVIDIA stops updating them given they do not
support more cards than 580.x does.
wrt bug #840439, potentially switching in dist-kernel will likely
require at least a warning somewhere or better yet a news item. Both
for old drivers (if still in-tree when this is done) or for users that
explicitly disabled fbdev=1 in their configs for one reason or another
(sometime it picks a bad resolution or cause other issues). Users that
cannot use or do not want fbdev=1 could probably use instructions to
revert with /etc/kernel/config.d and won't be able to use -bin anymore.
As for modeset=1, NVIDIA hasn't enabled that one by default "yet" but
realistically about every distros does. It is also needed for fbdev=1
to work and locking =1 behind USE=wayland makes that inconvenient.
--
Intentionally did this in a revbump *after* -r0 is stabilized so that
issues do not get mixed up with unrelated regressions in the driver
version itself. If users have -r1 issues not in -r0, they should know
to adjust their nvidia.conf as needed (a postinst warning is included).
Will stabilize this one a bit later if nothing major comes up.
Formerly meant to use a separate nvidia-580.conf but, given restored
USE=kernel-open in 590.x, nvidia-590.conf is fine for 580.x too.
Bug: https://bugs.gentoo.org/840439
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>
just skip the tests where namespace stuff is being run if the user
running the tests does not have subuid/subgid allocations
tests are automatically run in CI and as part of development (using an
account which has allocations)
Signed-off-by: Zen <z@pyl.onl>
Part-of: https://github.com/gentoo/gentoo/pull/45599
Signed-off-by: Michał Górny <mgorny@gentoo.org>
Or hopefully anyway, haven't attempted to reproduce to be spared long
builds but I'm almost certain that the fix is fine. This *used* to have
signal.h but was removed in cleanups years ago.
Hoping that there isn't further (different) issues with 2.43 that I
can't pickup without testing though.
Closes: https://bugs.gentoo.org/969495
Signed-off-by: Ionen Wolkens <ionen@gentoo.org>