FBO multisample error

[zed]

has anyone any idea whars going on here

http://www.zedzeek.com/junk/FBI_error.jpg

the above image is a FBO with multisample=4 samples
blitting with
glBlitFramebufferEXT(0, 0, W, H, 0, 0, W, H, GL_COLOR_BUFFER_BIT, GL_LINEAR );

if i use multisample=2 or disable multisample it renders perfectly
also if i bypass the FBO + render straight to the main window with multisample=4 it also renders perfectly

( i dont think its a driver bug, but ive stuffed up somewhere cause i seem to remember the same thing happened ~6 months ago )

ta zed

[namespace]

Have you tried to use GL_NEAREST instead of GL_LINEAR?
You don't scale your image during the blit, so it shouldn't change
the imagequality.

[zed]

cheers, GL_NEAREST makes no difference, as well as using smaller FBOs eg 640x480 (just in case it was a low memory issue)

[ruysch]

Perhaps you could give us the correct image ;-) its difficult to guess on your error with out having a correct version.

Whats your wrapping parameters ?

[zed]

4xAA + 2xAA
http://www.zedzeek.com/junk/error.jpg
GL_CLAMP_TO_EDGE or GL_REPEAT
perhaps its an error cause im using TEXTURE_RECTANGLE (ill try TEXTURE_2D )

*edit - texture_2d doesnt help out

[-NiCo-]

Now that you mention the wrapping modes...I've encountered this before too.
e.g. If I used a depth format of GL_DEPTH_STENCIL_NV and set the wrapping mode to GL_CLAMP_TO_EDGE everything got messed up.

[skynet]

Hi!
Just my two cents: I can confirm this wrong behaviour on a GF7900GTX (version 169.21) and on a QuadroFX 1400Go (169.39).
It is exactly as described: using 4xMS AA is causing the bug, 2xMS or AA disabled works as expected.
Using the same software with the same driver on a GF8800GTS doesn't show the problem, though (either 4x or other AA).
The only difference to OP's problem is, that in my case no render-to-texture is involved.