Red Hat Bugzilla – Bug 1015806
mythtv+nouveau+older card = playback double speed, audio plays normal speed
Last modified: 2015-01-09 03:21:51 EST
Description of problem:
I upgraded from Fedora 18 (nv driver) to 19 (nouveau driver, as 19 got rid of nv), and all mythtv frontend viewing started having double-speed video, but the audio was still 100% ok (normal speed). 18 to 19 did not change the myth versions.
I tried every myth option, every decoder, every deinterlacer, and it stayed double speed. Even setting audio to NULL didn't help. Did it on live TV as well as recordings.
When I did a while-watching options->playback->playbackdata it would show the mpg was 29.97 fps and the playback FPS was 59.75 to 59.99!!
mythavtest -O LocalHostName?=abcdefg /data/Mythtv/Recordings/1888_20130708085900.mpg
made no difference. It still played double-speed.
Testing with mplayer /data/Mythtv/Recordings/1888_20130708085900.mpg
and the video was fine (not double-speed). Problem is in myth only.
I switch to the binary nvidia driver and the video was fine (not double-speed) without any other changes. So the problem appears to be unique to the nouveau driver.
The output from myth while double-speed shows nothing out of the ordinary.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. boot and start X using nouveau
2. run mythfrontend
3. playback any recording
Plays video at double-speed. Audio is at normal speed.
Plays video and audio at normal speed.
Seems like it *might* be only on older 4-digit nvidia cards and not newer 3-digit cards based on other "me-too" reports. This may explain why it's not hitting more people.
See mythtv upstream bug:
Though I think this is probably a nouveau thing, not a myth thing.
I saw this with F19 and now see this with F20.
01:00.0 VGA compatible controller: NVIDIA Corporation GF106 [GeForce GTS 450] (rev a1)
[ 641.975] (--) NOUVEAU(0): Chipset: "NVIDIA NVC3"
This was fixed upstream and made it into Fedora's stock kernel sometime around 3.12 or 3.14.
See upstream bug link in comment in initial comment.
Kernel commit that fixed it was 7d3428cd4b2