| === jono is now known as Guest94995 | ||
| === lfaraone_ is now known as lfaraone | ||
| Dandel | ricotz, the recent packaging changes ( in xorg-edgers ) for nvidia 331.20 on precise breaks install... it does not install the nvidia-331-uvm package ( required since this is the kernel module ) | 16:06 |
|---|---|---|
| tseliot | he can reuse the packaging from precise-proposed | 16:13 |
| Dandel | tseliot, it looks like the issue is that the nvidia kernel module is not built and set to modprobe at boot | 16:13 |
| tseliot | Dandel: does the build fail? | 16:14 |
| Dandel | no | 16:14 |
| Dandel | the build, using this command works just fine... sudo dkms build -m nvidia -v 331-331.20 | 16:14 |
| Dandel | and it probes just fine also ( using the nvidia_331 module ) | 16:14 |
| Dandel | however the module was not loaded at boot ( extremely odd ) | 16:15 |
| tseliot | Dandel: yes, that's meant to be loaded and used on demand | 16:15 |
| tseliot | it's not something that the driver requires | 16:15 |
| Dandel | although I did a recent kernel update to linux-image-3.5.0-45-generic ( 3.5.0-45.68~precise1 ) on top of updating the nvidia driver | 16:16 |
| Dandel | where I updated the kernel *and* nvidia driver at the same time | 16:16 |
| Dandel | hmm... exact update order... Install nvidia-331 ( latest ) then install linux-image-3.5.0-45-generic | 16:18 |
| Dandel | so most likely the bug is not t he nvidia driver but the kernel package not attempting to build all dkms modules | 16:18 |
| Dandel | this is based on checking the dpkg.log file | 16:19 |
| tseliot | Dandel: make sure the headers for that kernel are installed too | 16:20 |
| Dandel | tseliot, They are installed... the particular log i was using is here ( http://paste.ubuntu.com/6685436/ ) | 16:22 |
| Dandel | line 1096 and 1484 to 1581 are of interest ( nvidia driver update and kernel update ) | 16:24 |
| Dandel | the tailing portion of the log is after I rebooted. | 16:24 |
| Dandel | brb... haft to restart xorg. | 16:32 |
| Dandel | tseliot, no go... it's definitely broken. | 16:38 |
| Dandel | the only thing even remotely listed is the dmesg log for nvidia is the alsa inputs ( kernel module does not load ). | 16:39 |
| Dandel | and xorg.0.log says it can't find the nvidia module | 16:39 |
| Dandel | tseliot, I'm forcing a package downgrade to see if that fixes the problem with the nvidia driver. ( from 331.20-0ubuntu8~xedgers~precise1 to 331.20-0ubuntu1~xedgers~precise1 ) | 16:47 |
| Dandel | hmm... that's annoying... can't use synaptic to purge the nvidia-331 packages... wants to force the install of nvidia-304 ><; | 16:52 |
| Dandel | I see... the bumblebee-nvidia package was what caused that 0o' | 16:55 |
| tseliot | Dandel: you can use my packages from precise-proposed | 17:03 |
| ricotz | Dandel, where is your bumblebee package coming from? | 17:04 |
| ricotz | the current bumblebee in trusty doesnt support 331, but precise doesnt even ship a package | 17:06 |
| ricotz | therefore there are updated saucy and trusty packages in edgers | 17:07 |
| Dandel | ricotz, the bumblebee ppa | 17:10 |
| Dandel | to be exact, this ppa ( https://launchpad.net/~bumblebee/+archive/stable ) | 17:10 |
| Dandel | ricotz, it's somewhat odd since I didn't have any issues until the absolute latest round of updates | 17:11 |
| ricotz | Dandel, ok, bumblebee got add to Recommends so it gets pulled in | 17:13 |
| Dandel | ricotz, I would also say that the opencl ICD ( and headers ) updates should get pulled into the latest lts. | 17:18 |
| ricotz | Dandel, yeah -- tseliot, i don't see a reason to revert the packaging split for precise? | 17:19 |
| Dandel | there is little to no reason to avoid including this since it's somewhat dumb to provide opencl headers package ( opencl-headers ) and have nothing to link it to... It is sort of on the order of allowing the opengl dev headers to install but not have any opengl libs. | 17:20 |
| tseliot | ricotz: what packaging split? | 17:20 |
| ricotz | tseliot, installing the opencl library in separate packages | 17:21 |
| tseliot | ricotz: we only do that in 14.04. I don't think I have introduced that change in precise | 17:22 |
| ricotz | tseliot, yes, that is why i am wondering what the reason for that was | 17:22 |
| Dandel | ricotz, there should be no reason to override the opencl lib. There is a set OpenCL ICD spec that exists where all you need to have is the nvidia driver (or ATI/AMD Driver) and the ocl-icd-libopencl1 package. | 17:22 |
| tseliot | ricotz: the packages in precise-proposed don't have that change | 17:23 |
| tseliot | ricotz: users might want to use those libraries without having to install the driver | 17:24 |
| Dandel | tseliot, where do i find the precise-proposed packages? | 17:25 |
| ricotz | tseliot, right, and they cant because the split isnt done for precise | 17:25 |
| tseliot | Dandel: in the precise-proposed repository | 17:25 |
| ricotz | tseliot, i am asking why it *isnt* done | 17:26 |
| ricotz | tseliot, this is most likely just calling for trouble on upgrades | 17:26 |
| ricotz | if there are not conflicts defined | 17:27 |
| tseliot | ricotz: in 14.04 you can't install all the drivers at the same time, whereas in 12.04 you can. This is why. | 17:28 |
| ricotz | hmm, still looks like a (maybe) broken upgrade path | 17:30 |
| Dandel | tseliot, does 14.04 enable vdpau on noveau and radeon drivers? ( a critical addition to the driver that was made in the last 6 months ) | 17:31 |
| tseliot | ricotz: nvidia will be replaced by a new nvidia which will also pull in the other libraries. It didn't break here | 17:32 |
| tseliot | Dandel: tjaalton mlankhorst should know about vdpau | 17:32 |
| Dandel | tseliot, I already mentioned this a couple of times... It's a new feature in the Mesa 10.0 tree | 17:33 |
| tseliot | we'll see then | 17:36 |
| Dandel | tseliot, There's also other improvements that also appear in mesa like Opencl ( since Mesa 9.0, it's the gallium state tracker called clover ) | 17:38 |
| tseliot | I think there was a discussion about opencl in debian | 17:39 |
| Dandel | Yea, and it definitely is because Mesa 10.0 added support for OpenCL ICD. | 17:42 |
| mlankhorst | it's not enabled for now | 18:08 |
| mlankhorst | as soon as it is clear that it becomes an improvement I may enable it, but I'm still reading about way too many vdpau related regressions | 18:13 |
| Dandel | mlankhorst, vdpau has some regressions but there is a lot of improvements by having it enabled... particularly for lower power machines. | 19:04 |
| Dandel | the only way for some machines to play 1080p video is to have vdpau enabled. A good example of one of these machines is the AMD E-series like the E350 or similar. | 19:06 |
| Dandel | also, there is major battery life improvements that can be seen on multiple machines like the amd APU based laptops | 19:06 |
| Dandel | hardware accelerated video decoding can easily cause major improvements in battery life for the effected laptops... There is a major difference between 2 hours of battery life and 3 hours of battery life. | 19:07 |
| Dandel | mlankhorst, what time period was the regressions listed? ( I know that there was a ton of issues initially, but most of the regressions/problems have been fixed ) | 19:10 |
| Dandel | mlankhorst, I believe the latest openelec development series ( generic configuration ) uses mesa with vdpau ( a single patch was added to enable vdpau playback with interlaced video ) | 23:11 |
| Dandel | ricotz, the precise-proposed 331 driver ( nvidia-331-updates ) worked... although it may be a good idea to change nvidia-settings package to have the version number with it ( to help avoid packaging avoid conflicts ) | 23:27 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!