375.82-7~bpo9+1

[Luca Boccassi]

On Wed, 2017-11-01 at 15:22 +0100, Andreas Beckmann wrote:
> On 11/01/2017 02:17 PM, Luca Boccassi wrote:
> > On Wed, 2017-11-01 at 13:25 +0100, Andreas Beckmann wrote:
> > > Hi Luca,
> > > 
> > > if you have some time, could you test 375.82-7~bpo9+1 as prepared
> > > in
> > > SVN?
> > > 
> > > Thanks
> > > 
> > > Andreas
> > 
> > Hi,
> > 
> > Sure can do, will have a go in the next couple days and report
> > back.
> 
> Reading the remaining bugs w.r.t. the provides again, I'm considering
> dropping all the remaining Provides of libglvnd package names and
> adding
> Breaks: lib*-glvnd-nvidia to nvidia-driver-libs (and maybe adding
> versioned Depends/Recommends on the corresponding libglvnd packages.
> That should hopefully result in an upgrade path switching to
> src:libglvnd only for regular driver installations.
> 
> I won't drop the packages right now, since we still need them for
> stretch and I want to keep the diff needed for stretch minimal.
> 
> The lib*-glvnd-nvidia packages will be still installable in
> sid/testing,
> just not along such things like X, mesa, or the nvidia driver.
> 
> 
> Andreas

That makes sense as a solution, too many problems with these multiple
incompatible implementations.

One issue, slightly related, is that it doesn't seem to be possible to
use libEGL without GLVND.

libegl1-nvidia, which is supposed to be the nonglvnd package, gets a
dependency on the following from shlibs:Depends at build time:

libglvnd0 | glvnd-nvidia-abi-375.20, libglvnd0 | libglvnd0-nvidia (>= 361.45)

That's because it's linked to libGLdispatch.so.0:

$ ldd /usr/lib/x86_64-linux-gnu/nvidia/current/libEGL.so.384.98
	linux-vdso.so.1 (0x00007ffc2671a000)
	libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007ff48e7af000)
	libGLdispatch.so.0 => /usr/lib/x86_64-linux-gnu/libGLdispatch.so.0 (0x00007ff48e4f9000)
	libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007ff48e15a000)
	/lib64/ld-linux-x86-64.so.2 (0x00007ff48ebb8000)


So it's not possible to completely remove libglvnd0 and use only non-
glvnd libraries, which is a problem for the bumblebee and optimus case.

Is there anything we can do about it? I'm guessing it must be a problem
with the upstream's build due to the linkage. This appears to be the
case back to the 375 series up to the very latest 384.

Kind regards,
Luca Boccassi
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: This is a digitally signed message part
URL: <http://lists.alioth.debian.org/pipermail/pkg-nvidia-devel/attachments/20171105/a2a4d1b4/attachment.sig>