[Etoys-notify] #7612 BLOC 8.2.0 (: XRender cursor rendering broken with XShmPutImage

[Zarro Boogs per Child]

#7612: XRender cursor rendering broken with XShmPutImage
-------------------------------+--------------------------------------------
   Reporter:  bert             |       Owner:  bernie                             
       Type:  defect           |      Status:  new                                
   Priority:  blocker          |   Milestone:  8.2.0 (was Update.2)               
  Component:  x window system  |     Version:  Development build as of this date  
 Resolution:                   |    Keywords:  8.2.0:? joyride-2208:- blocks:8.2.0
Next_action:  diagnose         |    Verified:  0                                  
  Blockedby:                   |    Blocking:  8008                               
-------------------------------+--------------------------------------------

Comment(by JordanCrouse):

 This seems to be the source of the issue - the damage function is
 "recovering" the pixmap by copying it from offscreen memory.  Then, when
 it tries to composite the mouse on a subsequent step, the src is already
 wrong.

 {{{
 <-- Unaccelerated copy from exaShmPutImage -->
 EXA fbCopyArea 33 66 129 130 -> 33 66
 EXA - DAMAGE REGION
 <-- Copy from damaged pixmap -->
 [EXA] - copy 0,0 [2c9a00] ->160,195 [61940] 43x59
 [EXA] - src = ffffffff dst = 55005500
 EXA - DAMAGE DONE
 <-- Copy src pixmap for OVER composite
     operation -->
 [EXA] - copy 160,195 [61940] ->0,0 [2c9a00] 43x59
 [EXA] - src = ffffffff dst = ffffffff
 <-- Composite operation -->
 [EXA]: SRC 2cb020,128,4,27,43
 [EXA]: DST 0,2048,2,1024,768
 }}}

 This definitely is related to XShm - the non-shared memory version in EXA
 goes down a different path.  In the above output, DAMAGE REGION and DAMAGE
 DONE wrap the function
 DamageDamageRegion() which is only called in the exa handler for shared
 memory PutImage.   At this point, its about a 50/50 chance that the bug
 still lives in the driver; we can't discount that EXA is doing the wrong
 thing, but up until now, every time I have blamed something on EXA, I have
 been proved wrong.  I'm going to investigate more, but until you hear
 otherwise, go ahead and turn off shared memory to resolve this blocker.
 Yeah, it will be a little bit slower, but correct is better then fast any
 day of the week.

-- 
Ticket URL: <http://dev.laptop.org/ticket/7612#comment:26>
One Laptop Per Child <http://laptop.org/>
OLPC bug tracking system