Created attachment 114863 [details] kernel.log IvyBridge Macbook with triple rotated monitors (--rotate left). TearFree enabled. Latest git kernel and xorg intel driver. Turn off screensaver/lock, then do: $ xset s blank $ xset s 5 Now wait for screen to blank, move mouse, wait for blank... At some point the monitors will unblank and the displays will be showing with no rotation (--rotate normal). Moving the mouse to a monitor and clicking on it seems to trigger a redraw back to the rotated state. Happens in XFCE and Gnome, kernel 3.16.0 and 4.0.0-rc6+. It is intermittent and sometimes difficult to trigger. On test in XFCE, it occurred 3 times out of 48. On Gnome test it seems easier to trigger, occurred 7/15 times. Think I managed to get Xorg and kernel logs at the right time.
Created attachment 114864 [details] xorg.log
Hmm the craziness in the other bug was due to bad interaction with the swcursor which should not apply here. :| However I can hope so it is worth confirming this is still a problem.
My guess is that the initial content supplied for switching on the CRTC doesn't represent the transformed view and the delay between the update to the real output is noticeable (even delayed until the next redraw far in the future).
Unfortunately the bug is still present. I guess the screensaver/screenlock usually trigger a full redraw of the screens which is why this only gets noticed when the user is using blank only. Odd that it happens more often in Gnome than XFCE though.
Ok, I noticed that it wasn't taking account of the rotation flag when choosing the initial contents for the scanout when waking up - so it was copying the frontbuffer without rotation onto the RandR shadow scanout before binding it to the CRTC. We did promptly try to redisplay the outputs (~16ms later but that is mostly due to the verbosity of the logging and CRTC enabling), but [ 2535.903] sna_mode_redisplay: shadow idle, skipping update so that means the unrotated content would persist until redraw. I believe I understand the problem correct (kudos for getting the right bit of the log!): commit e74ff581449e92d7933da1b5df6680cb880fc454 Author: Chris Wilson <chris@chris-wilson.co.uk> Date: Mon Apr 6 13:58:35 2015 +0100 sna: Force the TearFree redisplay when populating a rotated configuration We have to disable the "shadow idle; skipping" test if the CRTCs are showing stale content following a sna_crtc_attach(). Reported-by: Chris Bainbridge <chris.bainbridge@gmail.com> References: https://bugs.freedesktop.org/show_bug.cgi?id=89904 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Tested for 5 minutes with 5s between each blank and everything was perfect.
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.