Rendering in geode

[Huang, FrankR]

Thanks, Michel.
Actually this bug is caused by my new patch. My new patch is to handle the PictOpOver operation condition. So I still make sure this is caused by PictOpOver.
Does the PictOpOver has relationship with glyph rendering? If there is good client application to test the glyph rendering, that could be better. I don't know how to narrow down this bug. If other program can reproduce it, it is better. Any advice to this? Rendercheck ?

Frank

-----Original Message-----
From: Michel Dänzer [mailto:michel at daenzer.net] 
Sent: 2010年6月23日 16:23
To: Huang, FrankR
Cc: Jonathan Morton; xorg-driver-geode at lists.x.org; xorg-devel at lists.x.org
Subject: RE: Rendering in geode

On Mit, 2010-06-23 at 15:54 +0800, Huang, FrankR wrote: 
> 
> I found that right now the bug is in 1x1 source picture with repeat
> parameter. From my debug, there are two main condition for PictOpOver:
> 1)source: 1x1 repeatable(RGB565), mask region(A8), dest region(RGB565)
> 2)source: 1x1 repeatable(ARGB), mask region(A8), dest region(RGB565)
> For condition 1), I firstly convert the source to ARGB, then it is the
> same as condition 2). I use the HW registers gracefully handle this.
> But the side effect is that firefox has some characters missing. See
> link: 
> https://bugs.freedesktop.org/attachment.cgi?id=36432

As the same characters show the same corruption in different places, the
EXA glyph cache pixmap is probably getting corrupted.
exaGlyphCacheUploadGlyph() uses the UploadToScreen hook if applicable or
otherwise a PictOpSrc Composite operation for uploading glyphs to the
cache pixmap, you may want to check that the driver handles this
properly.


-- 
Earthling Michel Dänzer           |                http://www.vmware.com
Libre software enthusiast         |          Debian, X and DRI developer