A simple accelerated GL compositor
Felix Bellaby
felix at bellaby.plus.com
Sun Jul 9 18:46:30 PDT 2006
Tomasz Torcz wrote:
>> Just so that I am sure that we are on the same wavelength, could you try
>> setting up the fbc attributes as follows:
>>
>> line 1051:
>> int fbc_attr_mirror[] = {
>> GLX_DOUBLEBUFFER, False,
>> GLX_DEPTH_SIZE, 0,
>> GLX_RED_SIZE, 1,
>> GLX_GREEN_SIZE, 1,
>> GLX_BLUE_SIZE, 1,
>> GLX_ALPHA_SIZE, 1,
>> GLX_RENDER_TYPE, GLX_RGBA_BIT,
>> GLX_DRAWABLE_TYPE, GLX_PIXMAP_BIT,
>> None
>> };
>>
>> line 1146:
>> int fbc_attr_root[] = {
>> GLX_DOUBLEBUFFER, False,
>> GLX_DEPTH_SIZE, 16,
>> GLX_RED_SIZE, 1,
>> GLX_GREEN_SIZE, 1,
>> GLX_BLUE_SIZE, 1,
>> GLX_ALPHA_SIZE, 1,
>> GLX_RENDER_TYPE, GLX_RGBA_BIT,
>> GLX_DRAWABLE_TYPE, GLX_PIXMAP_BIT,
>> None
>> };
>
> Exits, hitting "return 1" in line 1166.
Okay, Tomasz, we were not on the same wavelength. It appears that any
attempt to create a GLXPixmap or GLXWindow with an alpha buffer results
in a failure on the mga drivers with the G550. Unfortunately, the
current code relies quite heavily on having access to alpha buffers.
Since this problem might well arise on other hardware as well, I shall
try out some work arounds for it, but it will take a bit of time. I am
afraid that you will have to wait until I post a new version of the
compositor before you can expect a positive result.
The debug listing that you posted is largely consistent with the
compositor going through the steps needed to draw the screen, but
failing to produce a visible image. However, there are a string of
XErrors (begining with the BadLength error and including the subsequent
BadDrawable errors) which are probably down to my failure to initialise
the copy_buffers variable to NULL. You might try correcting line 120 and
seeing whether the debug output looks cleaner.
Thank you for your patience in trying out a piece of graphics software
that fails to draw anything on your hardware. I might have been able to
make things a bit easier for you if my ISP had not decided to spend the
last few days making a complete hash of restructing their mail system.
Felix
More information about the xorg
mailing list