[elrepo] Centos 7.6 nvidia issue

Phil Perry phil at elrepo.org
Wed Dec 5 17:03:38 EST 2018


On 05/12/2018 21:11, Lamar Owen wrote:
> On 12/5/18 3:54 PM, Phil Perry wrote:
>> On 05/12/2018 18:43, Lamar Owen wrote:
>>>
>>> This change may be related to a new inability to compile MAME; it 
>>> errors out complaining about a missing symbol in libGL.so that is 
>>> supposed to be provided by the nvidia libraries.  I guess I should 
>>> just wait until the mirrors are updated.
>>>
>>
>> LibGL is now provided by the distro GLVND packages in el7.6 
>> (libglvnd-glx to be specific), although for development work I'm 
>> guessing you may also require libglvnd-devel to be installed which 
>> provides /usr/lib64/libGL.so to link against. 
> 
> Hi Phil, thanks for the reply.  Here's what I am getting:
> 
> [lowen at localhost ~]$ ldd -r /usr/lib64/libGL.so
>      linux-vdso.so.1 =>  (0x00007ffc263f5000)
>      libGLX.so.0 => /usr/lib64/nvidia/libGLX.so.0 (0x00007f43b4c51000)
>      libX11.so.6 => /lib64/libX11.so.6 (0x00007f43b4913000)
>      libXext.so.6 => /lib64/libXext.so.6 (0x00007f43b4701000)
>      libGLdispatch.so.0 => /usr/lib64/nvidia/libGLdispatch.so.0 
> (0x00007f43b442e000)
>      libdl.so.2 => /lib64/libdl.so.2 (0x00007f43b422a000)
>      libpthread.so.0 => /lib64/libpthread.so.0 (0x00007f43b400e000)
>      libc.so.6 => /lib64/libc.so.6 (0x00007f43b3c41000)
>      libxcb.so.1 => /lib64/libxcb.so.1 (0x00007f43b3a19000)
>      /lib64/ld-linux-x86-64.so.2 (0x00007f43b510d000)
>      libXau.so.6 => /lib64/libXau.so.6 (0x00007f43b3815000)
> undefined symbol: _glapi_tls_Current    (/usr/lib64/libGL.so)
> [lowen at localhost ~]$ rpm -qf /usr/lib64/nvidia/libGLdispatch.so.0
> nvidia-x11-drv-410.78-1.el7_6.elrepo.x86_64
> [lowen at localhost ~]$
> 
> That undefined symbol line is what is keeping MAME's build from finishing.
> 

Hmm. Once updated, my system shows:

$ ldd -r /usr/lib64/libGL.so
	linux-vdso.so.1 =>  (0x00007ffe9a9fb000)
	libGLX.so.0 => /lib64/libGLX.so.0 (0x00007f3b79116000)
	libX11.so.6 => /lib64/libX11.so.6 (0x00007f3b78dd8000)
	libXext.so.6 => /lib64/libXext.so.6 (0x00007f3b78bc6000)
	libGLdispatch.so.0 => /lib64/libGLdispatch.so.0 (0x00007f3b78910000)
	libdl.so.2 => /lib64/libdl.so.2 (0x00007f3b7870c000)
	libpthread.so.0 => /lib64/libpthread.so.0 (0x00007f3b784f0000)
	libc.so.6 => /lib64/libc.so.6 (0x00007f3b78123000)
	libxcb.so.1 => /lib64/libxcb.so.1 (0x00007f3b77efb000)
	/lib64/ld-linux-x86-64.so.2 (0x00007f3b795d4000)
	libXau.so.6 => /lib64/libXau.so.6 (0x00007f3b77cf7000)


$ ll /usr/lib64/libGL.so
lrwxrwxrwx. 1 root root 14 Dec  5 20:43 /usr/lib64/libGL.so -> 
libGL.so.1.7.0


$ rpm -qa nvidia-x11\*
nvidia-x11-drv-libs-410.78-2.el7_6.elrepo.x86_64
nvidia-x11-drv-410.78-2.el7_6.elrepo.x86_64


Please could you show 'rpm -qf /usr/lib64/libGL.so' and 'll 
/usr/lib64/libGL.so' - is it symlinking to the nvidia LibGL.so library? 
As you're still using the old nvidia packages, it should symlink to 
/usr/lib64/nvidia/libGL.so.1

But before we get too deep into this - does updating the nvidia packages 
fix the issue? If yum still isn't finding them, can you download and 
manually update?

I found this bug, maybe related:

https://bugzilla.redhat.com/show_bug.cgi?id=1440124



More information about the elrepo mailing list