On Fri, Mar 29, 2019 at 04:46:02PM -0500, Bruce Dubbs via blfs-dev wrote:
> >
> > And then, having discovered that my previous build had omitted
> > nouveau, I gave mesa a try using just the 'auto' settings.
> > The output from meson did not really indicate what it had enabled.
> >
> > Then I built it - added nouveau drivers, but SMALLER than my
> > previous build. Rerunning meson configure shows that 'osmesa' was
> > not enabled. I don't have valgrind installed, that would also be
> > auto. And gallium-nine is also disabled.
> >
> > But --sysconfdir seems to be redundant (it should be the default for
> > /usr, and anyway nothing gets installed in /etc by mesa-19).
> >
> > So, apart from the possibility of adding -Dbuild-tests=true, for a
> > full build of all available platforms and drivers, it appears that
> > there is no need to specify any of -Ddri-drivers=,
> > -Dgallium-drivers=, -Dplatforms=.
>
> Unless you want to be able to copy your binaries to another system. In my
> case my development system as a nvidia video card, but I sometimes copy
> things (in this case /opt/xorg) to my heat limited laptop that is an i7 with
> intel video.
>
> Am I reading this right?
>
> -- Bruce
>
I don't think you are. I've been doing this on an i7 haswell with
intel integrated graphics. My 'auto' build includes in /usr/lib
among others:
libvulkan_intel.so
libvulkan_radeon.so
libXvMCnouveau.so
In /usr/lib/dri I have
i965_dri.so
kms_swrast_dri.so
nouveau_dri.so
nouveau_drv_video.so
nouveau_vieux_dri.so
r200_dri.so
r300_dri.so
r600_dri.so
r600_drv_video.so
radeon_dri.so
radeonsi_dri.so
radeonsi_drv_video.so
swrast_dri.so
virtio_gpu_dri.so
vmwgfx_dri.so
In /usr/lib/vdpau I have
libvdpau_nouveau.so
libvdpau_r300.so
libvdpau_r600.so
libvdpau_radeonsi.so
(all the vdpau are symlinks to the .so.1.0.0 versions).
What I believe happens is that if you have provided all the
dependencies (e.g. LLVM's amdgpu part for amdgpu), and all the
platforms, then all of the gallium drivers, dri drivers, and
platform files (drivers, I suppose) get created.
In that case, provided you did not specify -march= with a value that
is not applicable to the machine you are copying *to*, I think
copying to a different machine should work.
AFAICS, all that changes is which drivers get compiled : specifying
them the way we currently do gives a small subset for modern
machines, letting 'auto' decide gives all of them.
But I'm coming to the conclusion that most people do not want to
build everything, and probably an option noting that "if those three
defines are not set then all drivers and platforms will be built"
is probably best.
ĸen
--
It is said that there are two great unsolved problems in computer
science: naming, cache invalidation, and off-by-one errors.
-- Ben Bullock
--
http://lists.linuxfromscratch.org/listinfo/blfs-dev
FAQ: http://www.linuxfromscratch.org/blfs/faq.html
Unsubscribe: See the above information page
