ati R580 - Aug 2009 & compiz still does not work for me

I see in the Xorg.0.log that AIGLX is turned off:

(**) Option "AIGLX" "off"
(II) Module glx: vendor="X.Org Foundation"
 compiled for 1.6.2, module version = 1.0.0
 ABI class: X.Org Server Extension, version 2.0
(**) AIGLX disabled
(II) AIGLX: Loaded and initialized /usr/lib/dri/swrast_dri.so

AIGLX is much better supported in -ati than -fglrx, so you should not need to disable it. In fact, this may be what is causing the sw rasterizer to be picked. Please test by moving aside your xorg.conf and restarting X, so it uses the built in defaults, and then attach your Xorg.0.log from that.

Desktop effects still dont work, but output of compize has changed:

compiz
Checking for Xgl: not present.
xset q doesn't reveal the location of the log file. Using fallback /var/log/Xorg.0.log
Detected PCI ID for VGA:
Checking for texture_from_pixmap: not present.
Trying again with indirect rendering:
Checking for texture_from_pixmap: present.
Checking for non power of two support: present.
Checking for Composite extension: present.
Checking screen 1Comparing resolution (1920x1200) to maximum 3D texture size (4096): Passed.
Checking for Software Rasterizer: Not present.
Checking for nVidia: not present.
Checking for FBConfig: present.
running under gnome seesion, checking for gnomecompat
Checking for Xgl: not present.
/usr/bin/compiz.real (core) - Error: Could not acquire compositing manager selection on screen 0 display ":0.0"
/usr/bin/compiz.real (core) - Fatal: No manageable screens found on display :0.0

(metacity:32582): atk-bridge-WARNING **: NOATBRIDGE: (null)
damage_screen (XSR): 1 rects, bounds: 0,0 (1920,1200)
expose_area (XSR): 1 rects, bounds: 1319,25 (601,821)
expose_area (XSR): 1 rects, bounds: 0,846 (1920,329)
expose_area (XSR): 1 rects, bounds: 0,1175 (1920,25)
expose_area (XSR): 1 rects, bounds: 636,0 (1284,25)
expose_area (XSR): 1 rects, bounds: 636,25 (683,379)
expose_area (XSR): 1 rects, bounds: 0,404 (1319,442)
expose_area (XSR): 1 rects, bounds: 0,0 (636,404)
repair_win (XSR): 1 rects, bounds: 0,0 (1920,1200)
repair_win (XSR): 1 rects, bounds: -9,1168 (1938,43)
repair_win (XSR): 1 rects, bounds: -9,-7 (1938,43)
configure notify 0 1 1
 extents (XSR): null
 xy (0 25), wh (1329 850)
no extents to damage !
resize_win (XSR): 1 rects, bounds: -9,18 (1347,868)
configure notify 0 1 1
 extents (XSR): 1 rects, bounds: -9,18 (1347,868)
 xy (0 25), wh (1329 850)
Inexplicable intersection with new extents!
resize_win (XSR): 1 rects, bounds: -9,18 (1347,868)
configure notify 0 1 1
 extents (XSR): 1 rects, bounds: -9,18 (1347,868)
 xy (0 25), wh (1329 850)
Inexplicable intersection with new extents!
resize_win (XSR): 1 rects, bounds: -9,18 (1347,868)
configure notify 0 1 1
 extents (XSR): 1 rects, bounds: -9,18 (1347,868)
 xy (0 25), wh (1329 850)
Inexplicable intersection with new extents!
resize_win (XSR): 1 rects, bounds: -9,18 (1347,868)
repair_win (XSR): 1 rects, bounds: -109,-107 (19,19)
configure notify 0 1 1
 extents (XSR): null
 xy (0 25), wh (646 433)
no extents to damage !
resize_win (XSR): 1 rects, bounds: -9,18 (664,451)
configure notify 0 1 1
 extents (XSR): 1 rects, bounds: -9,18 (664,451)
 xy (0 25), wh (646 433)
Inexplicable intersection with new extents!
resize_win (XSR): 1 rects, bounds: -9,18 (664,451)
configure notify 0 1 1
 extents (XSR): 1 rects, bounds: -9,18 (664,451)
 xy (0 25), wh (646 433)
Inexplicable intersection with new extents!
resize_win (XSR): 1 rects, bounds: -9,18 (664,451)
configure notify 0 1 1
 extents (XSR): 1 rects, bounds: -9,18 (664,451)
 xy (0 25), wh (646 433)
Inexplicable intersection with new extents!
resize_win (XSR): 1 rects, bounds: -9,18 (664,451)
configure notify 0 1 1
 extents (XSR): 1 rects, bounds: -9,18 (664,451)
 xy (5 49), wh (646 433)
Inexplicable intersection with...

Awesome thanks. What's going on here is that previously when you had AIGLX off, the system was using software rendering, which works but is very slow (and probably processor intensive) since it is all done in software. Compiz refuses to run with software rendering just as a matter of principle, so that's why it wasn't working. 3D games may work but will be *very* slow.

Now, you've got hardware acceleration enabled, but for some reason it is buggy. Now the compiz wrapper script actually allows the compiz.real binary to start, but it craps out due to a (real) driver bug. You may find other 3D applications also fail during startup.

The good news is now we have something we can forward upstream.

In chatting with upstream we've found a workaround and also identified it as not an X bug but rather a metacity bug that compiz is not working around properly.

<agd5f> asac: accoridng to google others fixed that problem by disabling metacity compositer
 gconftool-2 --type bool --set /apps/metacity/general/compositing_manager false
<asac> rock & roll ;) thanks guys
<otaylor> bryce: metacity bug, actually.
<Sarvatt> watch out for this bug if you're calling something with --replace on karmic with metacity the default WM asac: https://bugs.edge.launchpad.net/ubuntu/+source/metacity/+bug/389686
<otaylor> bryce: when it's asked to unmanage the screen it sets the selection owner of the compositor selection to None with a timestamp of CurrentTime, which is newer than the timestamp that compiz is trying to use to replace the window manager and compositing manager selections
<bryce> otaylor, wild, ok thanks.
<otaylor> See section 2.3.1 of the icccm for the specific recommendation that to give up ownership of a selectio you should use the timestamp you used to acquire it
<otaylor> bryce: http://bugzilla.gnome.org/show_bug.cgi?id=530702

How is this a bug in compiz, exactly? Unlike other problems compiz might have with other programs metacity can be fixed and it should be pretty easy to do so. The only thing I can think of compiz could do here is use a timestamp sometime in the distant future but that hardly seems proper.

It looks like at this point this bug is actually the same as bug 178953 which I've actually got a fix for. Marking as dupe.