Not sure if bugs in RenderAccel (which is clearly marked as experimental) are fair game for bug reports, but here goes.
Attached a screenshot of corruption i'm seeing in an xterm with RenderAccel on. The corruption doesn't happen with RenderAccel off.
Software: 8174 drivers, Xorg 6.8.2, metacity 2.12.1, Keith Packard's hacked xterm to draw with translucent background (found at http://freedesktop.org/~keithp/xterm-trans.tar.gz)
, running xcompmgr. GLX is disabled. All programs are built 64-bit, from Ubuntu's breezy distro.
Hardware: Athlon64 3700+ (San Diego), Gigabyte GA-K8N51GMF-9 motherboard with onboard GeForce 6100 chip, set to use 64MB for video.
To reproduce the problem: I started up the hacked xterm and typed "asdfasdfasdfasdf" over and over. If I move, resize, or otherwise make the xterm redraw the corrupted parts, they reappear uncorrupted.
The same thing sometimes happens with the stock Xorg xterm too, if I start the xterm and do not move it from its initial position. Once I move the stock xterm, the problem never happens on that window. But on the Keith's xterm it happens all the time.
Again, the problem doesn't appear when RenderAccel is off, but then of course composite + xcompmgr are dog slow. I'm not 100% clear on how Render works, but I'm pretty surprised that the screenshot was able to capture the corruption.
This used to work fine on roughly the same software way back when I was running an AGP Geforce3 Ti on an Athlon 2500+ with the 7676 drivers. I don't have any other hardware to test this with, unfortunately.
Thanks in advance, and I hope this helps the driver writers. This is my one missing piece of eye candy.