Created attachment 16185 [details] My Xorg log X.org version: 1.4.0.90 Kernel version: 2.6.24 Driver version: 1.2.1, commit 6f34607b The display blanks for a second or so when I start or stop watching video using mplayer. Sometimes the blanking is hardly noticeable, but on other times it is fairly annoying. I have a X1400 card. When this occurs, the logs do not show anything out of the ordinary. This also did not happen in earlier git builds. I've yet to determine at which point the regression happened.
We don't do anything special for video playback. There is not Xv extension in the driver, yet. You may want to try out if turning off XAA helps there. Put an 'Option "AccellMethod" "ShadowFb" into the device section of your config file. I would assume that this problem has something to do with the introduction of XAA. If above test doesn't help please do a git-bisect to determine where this problem was introduced. Assigning to reporter for feedback. Please assign back when done.
Created attachment 16214 [details] Bisect results
Unfortunately setting Option "AccelMethod" "ShadowFb" did not help. I was able to do a bisect though.
Cool, thanks for the bisect! This indicates that mplayer is using DPMS otherwise the reset function for the output wouldn't be touched.
Glad I could help. Git seems to be a handy tool - bisecting was a breeze. Can you figure out some resolution for this? Or is it even fixable?
mplayer inhibits the screensaver when you're playing video with it; I suspect that's why it touches DPMS.
this looks like the same issue as bug 14869
Could be. Just to verify your suspicions. I ran "xset -dpms" before mplayer and no blanking. The action of disabling DPMS causes the aforementioned effect though.
Created attachment 16221 [details] [review] Patch for testing. Thanks for testing! I've attached a patch which adds a. more detailed debugging output for '-logverbose 7' mode. b. attempts to fix this problem. The change which caused the problem adds a full output pll reset and data resync. This would not be required in the POWER_RESET state when only the output line drivers are off. The old code did the reset in the Set() code, however it is required whenever we return from POWER_SHUTDOWN and there is nothing that would guarantee that there is always a call to Set() before. If you don't mind, could you please test this patch for me? Thanks!
Sure. I ran the patch and X starts just fine, but with the display blanked. I'll attach my log.
Created attachment 16222 [details] Xorg log after the patch
Created attachment 16224 [details] [review] Additional patch. I think I know where the problem is. Could you please try this patch on top?
Great! It seems to work. No more blanking with mplayer. I started and stopped playback a few times as shown in the log after "SetClientVersion: 0 9".
Created attachment 16226 [details] Log after the additional patch
Cool! Thanks for testing! Fix committed.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.