Bug#879261:
Thank you Luca - I did see that the -nvidia libs are being dropped on the -8 changelog but I had no idea what that actually meant until I read your explanation. I was assuming these were modified library versions by Nvidia with enhancements to make the most out of the hardware but I guess their removal goes to show they're equivalent in any way. In that sense, my system should be fine right now, with maybe just a little headache to make sure all -nvidia packages are replaced properly in the next update. Thanks again for the extra information! On 5 November 2017 at 15:09, Luca Boccassi wrote: > On Sun, 2017-11-05 at 00:03 -0200, Alex Henry wrote: > > Just wanted to apologize for my last message since I missed the > > "Fixed in > > version nvidia-graphics-drivers/375.82-8" message in the header. It > > also > > only seems to have been marked as a resolved bug (bugtracker > > category) as I > > was writing it. I'll wait for 375.82-8 to trickle down to testing and > > try > > to install it again and report here. > > > > Also, usually the bug tracker adds an internal message to the report > > once > > the bug is resolved, not sure why it didn't happen here? Maybe it > > doesn't > > work like that anymore... > > > > Anyway, thanks for the fix, looking forward to having my nvidia > > driver > > fully functional again soon! > > Hi, > > What you noted in the previous email is correct - the problem appears > to be due to having both i386 and amd64 libglvnd0-nvidia. > > There are other problems with that package as noted elsewhere - namely > not all symbols are exposed, for whatever reason. Although it's an open > source library we can't know since it's distributed as a binary blob, > we don't know how Nvidia builds it. > > libglvnd0 is the equivalent - they are built from the same source so > the -nvidia specific ones are getting deprecated and should be removed > in favour of the "system" ones in -8. > > Kind regards, > Luca Boccassi
Bug#879261:
On Sun, 2017-11-05 at 00:03 -0200, Alex Henry wrote: > Just wanted to apologize for my last message since I missed the > "Fixed in > version nvidia-graphics-drivers/375.82-8" message in the header. It > also > only seems to have been marked as a resolved bug (bugtracker > category) as I > was writing it. I'll wait for 375.82-8 to trickle down to testing and > try > to install it again and report here. > > Also, usually the bug tracker adds an internal message to the report > once > the bug is resolved, not sure why it didn't happen here? Maybe it > doesn't > work like that anymore... > > Anyway, thanks for the fix, looking forward to having my nvidia > driver > fully functional again soon! Hi, What you noted in the previous email is correct - the problem appears to be due to having both i386 and amd64 libglvnd0-nvidia. There are other problems with that package as noted elsewhere - namely not all symbols are exposed, for whatever reason. Although it's an open source library we can't know since it's distributed as a binary blob, we don't know how Nvidia builds it. libglvnd0 is the equivalent - they are built from the same source so the -nvidia specific ones are getting deprecated and should be removed in favour of the "system" ones in -8. Kind regards, Luca Boccassi signature.asc Description: This is a digitally signed message part
Bug#879261:
Just wanted to apologize for my last message since I missed the "Fixed in version nvidia-graphics-drivers/375.82-8" message in the header. It also only seems to have been marked as a resolved bug (bugtracker category) as I was writing it. I'll wait for 375.82-8 to trickle down to testing and try to install it again and report here. Also, usually the bug tracker adds an internal message to the report once the bug is resolved, not sure why it didn't happen here? Maybe it doesn't work like that anymore... Anyway, thanks for the fix, looking forward to having my nvidia driver fully functional again soon!
Bug#879261:
I have also recently encountered problems with the nvidia-driver-libs dependencies on previous updates in my buster system (Debian testing) but somehow managed to work around it by using one of the alternative dependency options: I am currently using libgl1-nvidia-glx (instead of libgl1-nvidia-glvnd-glx) and libegl1-nvidia (instead of nvidia-egl-icd). I tried figuring out the root of the issue at the time so as to report a bug here but really couldn't since the dependency tree for the driver's libraries is quite complex and eventually just went with the "alternative options" route instead, through the aforementioned packages. However, when trying to upgrade my system today, I have come across an issue that is much more punctual and thus easier to report, which might have been the core of the problem all along (or not, I don't know). The root problem is that even though libglvnd0-nvidia indicates that it replaces libglvnd0, apt somehow doesn't seem to take that into account. Trying to remove both amd64 and i386 versions of libglvnd0 and attempting to install the two version of libglvnd0-nvidia instead, which should be a simple and straightforward operation since the second claims to replace the first, instead causes 7 broken dependencies on my system (seen via the aptitude package manager) - among which is libegl1, for example. I am not sure why this is happening but my hunch would be that libglvnd0-nvidia and libglvnd0-nvidia:i386 state that they both "replace" AND "conflict" with their standard libglvnd0 versions. I imagine this could confuse apt because it would then assume that "this package cannot ever be installed because it conflicts with a package used by so many other programs", which in turn leads it to ignore the actual "replaces" field. Anyway, whatever the real cause of the problem is, the fact is that, for all practical purposes, libglvnd0-nvidia is not able to replace libglvnd0 in my system - and the same for their i386 versions. If the maintainers could please look into this issue and try to solve it, it would be much appreciated, since, of course, I would want to have the nvidia version of the library installed since I assume it is enhanced to make use of the accelerated hardware. Also as it stands, libglvnd0-nvidia seems to be completely worthless as a package right now, since it cannot be installed, which for me sounds like a grave problem (as far as the Debian bug reporting guidelines define "grave": "makes the package in question unusable or mostly so"). Thank you for the excellent nvidia support on Debian and I'm certainly happy that I can currently keep using the non-nvidia library version (instead of just having the driver break entirely) - but please consider looking into this issue (even if only to realize it's an apt bug instead, which we should also report, if that's indeed the case). Let me know if I can help in any other way. Lastly, this bug is breaking routine upgrading in multiple system thats nvidia-driver (mine's and the original reporter's, at the very least) so please consider that this might be affecting a big number of Debian users who rely on these official Debian packages.
Bug#879261: libglvnd0-nvidia: Upgrade is broken
On 10/21/2017 11:39 AM, Marcus Lundblad wrote: > Since a few days, upgrading the system has stopped working (in buster) > > Trying to run sudo apt upgrade now shows a broken packages message: > > marcus@gemini:~$ LANG=C sudo apt upgrade > The following packages have unmet dependencies: > libglvnd0-nvidia : Conflicts: libglvnd0 > libglvnd0-nvidia:i386 : Conflicts: libglvnd0 > E: Broken packages > marcus@gemini:~$ since libglvnd0-nvidia is not a drop-in replacement for libglvnd0 due to some missing symbols, please install libglvnd0 and remove libglvnd0-nvidia Andreas