Motion blur and Water effect plugins are incompatible

[Saivann Carignan]

When running the water rain effects with motion blur enabled, windows become sunny light yellow and keep flashing after each water drop. This is a very strange behavior.

Steps to reproduce :

1. Enable Motion blur and Water effects
2. Start Motion blur with CTRL + F12
3. Start the rain with Shift + F9

This bug has been tested under Linux Ubuntu Hardy with compiz 1:0.7.0-0ubuntu3
nVidia Corporation GeForce 8300 GS (rev a1) with nvidia-glx-new proprietary drivers.

[TerryG (tgalati4)]

Triaged to Confirmed. On Gutsy with Intel945 chipset, having both effects causes a tremendous slowdown in desktop response, but no graphic weirdness. Rain effect by itself is speedy. Adding Motion Blur slows it down with the following terminal messages:

A handler is already registered for the path starting with path[0] = "org"
A handler is already registered for the path starting with path[0] = "org"
A handler is already registered for the path starting with path[0] = "org"
A handler is already registered for the path starting with path[0] = "org"
A handler is already registered for the path starting with path[0] = "org"
A handler is already registered for the path starting with path[0] = "org"

[Daniel Stiner (danstiner)]

I am unable to reproduce the yellowed window effect by turning on rain and motion blur in Hardy Heron Beta.
GeForce 8600 GT with nvidia-glx-new and compiz 1:0.7.4-0ubuntu5.
Though I do get many of the handler is already registered messages if I run compiz in a terminal.

[Saivann Carignan (oxmosys)]

I also confirm that the situation improved! The only visible problem that I get when running these two plugins simultaneously is that the top of the windows never stop flashing between pure orange, then yellow and finally it become completely white.

[Daniel Stiner (danstiner)]

Is your "Motion Blur Mode" set to Texture Copy or Accumulation buffer?
I get absolutely no weird effects with Accumulation buffer (which is the default one), but the less resource intensive Texture Copy mode does exactly what you describe. Everything turns a shade of white, and the window borders are mainly a yellowish color, but flash bright orange (human theme orange) Sometimes the whole screen even flashes its normal colors for a fraction of a second, and then goes back to the whitish shade.

[Saivann Carignan (oxmosys)]

I confirm what you say, Texture copy is the only one that cause such problems, Accumulation buffer works great, but did not seem to be the default one here in my case, I might be wrong.. Using nvidia geforce 8300 with restricted nvidia drivers.

[Daniel Stiner (danstiner)]

Well, it must have been my settings that were messed up, after disabling motion blur, and re-enabling, the default is now texture copy. Very odd.
Now that this bug has been confirmed for texture copy mode only, I commented on the Compiz Bugzilla page:
http://bugs.opencompositing.org/show_bug.cgi?id=808

[Saivann Carignan (oxmosys)]

Still confirmed on intrepid alpha 5

[Daniel Stiner (danstiner)]

So now what do we do?
I voted for the bug upstream, but nothing seems to be getting done.

[Saivann Carignan (oxmosys)]

I don't have necessary skills to work on this, so I guess that all I can do so far is to keep the bug report updated.

[Sam Spilsbury (smspillaz)]

This sounds a /lot/ like a driver based issue to me. I don't see any reason as to why ARB_fragment_program and the simple texture copy and replace code should interfere here, unless there are bugs in compiz' opacity handling. Although, I can say that I have an 8600GT and am using 191.x and am seeing this issue as well (although, it is a general issue with the texture copy logic and doesn't just happen when water is enabled [but it does when blur is enabled too])

Could somebody please triage this with a FOSS driver that is known to support fragment programs and is relatively fast? I'm thinking r4xx/r3xx.

Cheers,

Sam

[Sam Spilsbury (smspillaz)]

BTW, the 'handler already registered' messages are harmless and can be ignored, that is just DBUS trying to set up another connection that already existed for another compiz instance.

[jhfhlkjlj (fdsuufijjejejejej-deactivatedaccount)]

This was indeed an old driver bug with nvidia and has since been fixed (can't find the old bug # otherwise I'd just link as dupe).

Regardless, this is an old bug for an unsupported version of compiz. If this issue still persists, please open up the report again.