Re: Flash/Mozilla/Composite bug details?

Around 10 o'clock on Aug 13, Alan Coopersmith wrote:

> Can someone send an explanation between now and Monday of what the
> bug in Flash is and what they need to do with the way they handle
> visuals to fix it?   (Perhaps even a simple test case - is it just
> build the CVS head, turn on composite and try to run Flash?)

The test case is simple -- run mozilla against an X server with the ARGB 
visual to a site using flash ( http://www.kimpossible.com is my daughter's 
favorite at the moment)

The sympotom is an attempt by Flash to use PutImage with a depth 32 image
to a 16 bit window.

Here's a bit of an xscope trace:

         ............REQUEST: PutImage
             sequence number: 0000009e
                      format: ZPixmap
              request length: ff56
                    drawable: DWB 02800001
                          gc: GXC 0280001a
                            function: Copy
                          plane-mask: ffffffff
                          foreground: 00000000
                          background: 00000001
                       width: 02f8
                      height: 0056
                       dst-x: 0
                       dst-y: 0
                    left-pad: 00
                       depth: 20
         ............REQUEST: GetInputFocus
             sequence number: 0000009f
              request length: 0001
388.21:                                           64 bytes <-- X11 Server 7
                                         ..............ERROR: Match
                                             sequence number: 009e
                                                minor opcode: 0000
                                                major opcode: 48

And here's a stack trace from mozilla:

#41 0x404ffd11 in gdk_x_error (display=0x0, error=0x40965edc)
    at gdkmain-x11.c:530
#42 0x40692d64 in _XError (dpy=0x8838cf0, rep=0xbfffb2b0) at XlibInt.c:2909
#43 0x406912e3 in _XReply (dpy=0x8838cf0, rep=0xbfffb2b0, extra=0, discard=1)
    at XlibInt.c:1839
#44 0x4068c3b4 in XSync (dpy=0x8838cf0, discard=0) at Sync.c:45
#45 0x4068c475 in _XSyncFunction (dpy=0x0) at Synchro.c:34
#46 0x406853a5 in XPutImage (dpy=0x8838cf0, d=41943041, gc=0x868b6e8,
    image=0x868b760, req_xoffset=0, req_yoffset=0, x=0, y=0, req_width=760,
    req_height=86) at PutImage.c:1037
#47 0x4614f085 in PlatformBitBuffer::BltToScreen ()
   from /usr/lib/flashplugin-nonfree/libflashplayer.so
#48 0x4611eabd in CorePlayer::DrawScreen ()
   from /usr/lib/flashplugin-nonfree/libflashplayer.so
#49 0x461973cb in PlatformPlayer::NsSetWindow ()
   from /usr/lib/flashplugin-nonfree/libflashplayer.so
#50 0x4619b34e in NPP_SetWindow ()
   from /usr/lib/flashplugin-nonfree/libflashplayer.so
#51 0x46199580 in Private_SetWindow ()
   from /usr/lib/flashplugin-nonfree/libflashplayer.so
#52 0x42149127 in ns4xPluginInstance::InitializePlugin ()
   from /usr/lib/mozilla-firefox/components/libgkplugin.so

And here's where that image is created:

#0  XCreateImage (dpy=0x8850658, visual=0x86a7868, depth=32, format=2,
    offset=0, data=0x1f <Address 0x1f out of bounds>, width=728, height=31,
    xpad=32, image_bytes_per_line=0) at ImUtil.c:328
#1  0x44f8090c in PlatformBitBuffer::CreateXImage ()
   from /usr/lib/flashplugin-nonfree/libflashplayer.so
#2  0x44f80d2d in PlatformBitBuffer::CreateScreenBits ()
   from /usr/lib/flashplugin-nonfree/libflashplayer.so
#3  0x44f80622 in PlatformBitBuffer::PlatformBitBuffer ()
   from /usr/lib/flashplugin-nonfree/libflashplayer.so
#4  0x44f5134e in CorePlayer::UpdateBuffer ()
   from /usr/lib/flashplugin-nonfree/libflashplayer.so
#5  0x44f50a43 in CorePlayer::DrawScreen ()
   from /usr/lib/flashplugin-nonfree/libflashplayer.so
#6  0x44fc93cb in PlatformPlayer::NsSetWindow ()
   from /usr/lib/flashplugin-nonfree/libflashplayer.so
#7  0x44fcd34e in NPP_SetWindow ()

The specified visual is the ARGB visual and was selected by the flash 
player which appears to have a fixed list of formats to try and doesn't 
appear to know which visual is associated with the output window.

Here's where the visual was selected:

#0  XMatchVisualInfo (dpy=0x8873268, screen=0, depth=32, class=4,
    visual_info=0xbfffb430) at VisUtil.c:202
#1  0x4633aa80 in UnixCommonPlayer::ChoosePixmapFormat ()
   from /usr/lib/flashplugin-nonfree/libflashplayer.so
#2  0x4637d373 in PlatformPlayer::NsSetWindow ()
   from /usr/lib/flashplugin-nonfree/libflashplayer.so
#3  0x4638134e in NPP_SetWindow ()
   from /usr/lib/flashplugin-nonfree/libflashplayer.so
#4  0x4637f580 in Private_SetWindow ()
   from /usr/lib/flashplugin-nonfree/libflashplayer.so
#5  0x42149127 in ns4xPluginInstance::InitializePlugin ()
   from /usr/lib/mozilla-firefox/components/libgkplugin.so

As we can see, it's looking for a depth 32 TrueColor visual, which does 
not match the visual for the window into which this image will be put.

I suspect what they should be doing is using GetWindowAttributes to 
discover the visual for their window and using that instead of attempting 
to get at it via this round-about route.

-keith

_______________________________________________
release-wranglers mailing list
[email protected]
http://freedesktop.org/mailman/listinfo/release-wranglers

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.5 (GNU/Linux)
Comment: Exmh version 2.3.1 11/28/2001

iD8DBQFBHR+DQp8BWwlsTdMRAlS9AJ9YlQ2fe68OGvvsw1aoLviA/B0ztwCcCGn8
LIDYjhd0q0CKuoWwqChpzns=
=sRVt
-----END PGP SIGNATURE-----