[Intel-gfx] [PATCH] i915: Do not force the incoming source alpha to 1.

Eric Anholt eric at anholt.net
Tue Nov 10 20:30:06 CET 2009


On Tue, 2009-11-10 at 15:52 +0000, Chris Wilson wrote:
> RENDER specifies that textures are extended with the clear colour in the
> EXTEND_NONE case. By forcing the border colour to contain 1, we were
> padding with black instead. This is a long standing bug that causes
> numerous failure in cairo and has been frequently bemoaned about in
> bugzilla.
> 
>   https://bugs.freedesktop.org/show_bug.cgi?id=17933
>   x8r8g8b8 doesn't sample alpha=0 outside surface bounds
> 
> However, one consequence of this is that we now must trust the incoming
> source alpha for x8r8g8b8 surfaces (and similar ignored alpha formats).
> We workaround this by ensuring that we fill the outgoing alpha channel
> with 1 for such targets, but we must also rely on the X server to fill
> the alpha channel on incoming RGB images.

The server doesn't fill the alpha channel on incoming rgb images,
though.  And can't you take the same pixmap and use it for both x8r8g8b8
and a8r8g8b8?

-- 
Eric Anholt
eric at anholt.net                         eric.anholt at intel.com


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part
URL: <http://lists.freedesktop.org/archives/intel-gfx/attachments/20091110/7ef117df/attachment.pgp>


More information about the Intel-gfx mailing list