Created attachment 73613 [details]
the bug

Hmm, uninitialised data or bad damage. Can you share the java app?

(In reply to comment #2)
> Hmm, uninitialised data or bad damage. Can you share the java app?

Sure, it's josm:
http://josm.openstreetmap.de/

Created attachment 73614 [details]
another corruption, only backgrounds of some texts

Slight snag... After the splash screen completes and josm finishes loading, I'm faced with a blank application, using UXA or SNA. Weird.

Well, I can see the text corruption on pnv + unity, josm/java=1.6. It reminds me of some of the earlier damage bugs.

Great. The machine I saw it died. And I can't reproduce at all on another pnv (java6/java7, bare X/gnome/etc).

Jiri, can you confirm this problem still exists on current -intel.git? (I can be optimistic!)

Hmm, I have resurrected that machine and have been able to reproduce the text corruption (whilst scrolling a list). On that machine it is only reproducible whilst running under unity.

Seems to be something to do with:

diff --git a/src/sna/sna_accel.c b/src/sna/sna_accel.c
index 2cc45e4..163d4c5 100644
--- a/src/sna/sna_accel.c
+++ b/src/sna/sna_accel.c
@@ -1744,7 +1744,7 @@ done:
                if ((flags & MOVE_ASYNC_HINT) == 0) {
                        DBG(("%s: syncing CPU bo\n", __FUNCTION__));
                        kgem_bo_sync__cpu_full(&sna->kgem,
-                                              priv->cpu_bo, flags & MOVE_WRITE)
+                                              priv->cpu_bo, 1);//flags & MOVE_W
                        assert(!priv->shm || !kgem_bo_is_busy(priv->cpu_bo));
                }
                if (flags & MOVE_WRITE) {
@@ -2267,7 +2267,7 @@ out:
        if ((flags & MOVE_ASYNC_HINT) == 0 && priv->cpu_bo) {
                DBG(("%s: syncing cpu bo\n", __FUNCTION__));
                kgem_bo_sync__cpu_full(&sna->kgem,
-                                      priv->cpu_bo, flags & MOVE_WRITE);
+                                      priv->cpu_bo, 1);//flags & MOVE_WRITE);
        }
        priv->cpu = (flags & MOVE_ASYNC_HINT) == 0;
        assert(pixmap->devPrivate.ptr);

My version of the bug at least is only present in the Ubuntu kernel - installing dinq+ fixes the scrolling through the text boxes (afaict).

Created attachment 74775 [details]
bug in firefox

Is this the same bug?

Yes I would say that was the same bug. But I don't think it is the same bug as I am seeing.

Another bug worth testing with -intel master, following:

commit c6a59bee3bf5f42147f75a714f2b9aa26590329e
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Wed Mar 6 19:01:50 2013 +0000

    sna: Disable read-read optimisations
    
    There is still a lurking issue, so punt on the optimisation for now.
    
    Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=61628
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

Created attachment 76270 [details]
persisting bug in FF

(In reply to comment #13)
> Another bug worth testing with -intel master, following:

Ok, with 2.21.3-90-g93ecd56, the java issue is gone, it seems. But I still see this with firefox as can be seen in the attachment. There are two glitches: window decoration and image on a page marked as BAD in the picture.

The only window title broken is of firefox though. Other window titles are rendered OK. Switching effects off and on made the image misrenrendering go away, so I also put the picture of how should look like with effect enabled.

Hmm. Incoherence along PutImage, or something even more bizarre. :|

The bug can't be reproduced on G45 machine. Our test environment as following:
Mesa:     (9.1)f00ae9c773dc944f0d900ef5c50c3b417ad95e84
Xserver:  (server-1.13-branch)xorg-server-1.13.4
Xf86_video_intel:(master)2.21.7-5-g87295b1ef85505689ce326137c2794230fb3f35f
Cairo:(master)631bf299256e11a17511977f357e0353fb5615f7
Libva_intel_driver:(master)31caada2967b94705d78ab7f6d07965ad7f13d42
Kernel:(drm-intel-fixes) 3598706b52cb45ba0a9e8aa99ce5ac59140f2b8b
Os:     Fedora17

This could be the coherency issues with the CPU data fixed now in -intel.git. Any chance you are still able to reproduce and test?

(In reply to comment #17)
> This could be the coherency issues with the CPU data fixed now in
> -intel.git. Any chance you are still able to reproduce and test?

Yeah, it's still an issue. Immediately after KDE start, whole screen flickered several times with horizontal lines.

(In reply to comment #18)
> (In reply to comment #17)
> > This could be the coherency issues with the CPU data fixed now in
> > -intel.git. Any chance you are still able to reproduce and test?
> 
> Yeah, it's still an issue. Immediately after KDE start, whole screen
> flickered several times with horizontal lines.

Disregard this comment...

Created attachment 80879 [details]
issue on xterm's decoration

It's still broken with 2.21.9-28-g7ebcf94.

(In reply to comment #20)
> Created attachment 80879 [details]
> issue on xterm's decoration
> 
> It's still broken with 2.21.9-28-g7ebcf94.

But this is rather my bug 63715. I think this issue is gone.