1. Doesn't happen if the texture is mipmapped.
2. Doesn't happen if the texture is pre-compressed.
3. Doesn't happen if the internal format is uncompressed.
So we only see this on non-mipped, driver-compressed RGBA images...
cheers
ben
Dario Accornero wrote:
Ben,
I can confirm this is totally broken in 10.5.3 -- doing PCF for shadows
on a Radeon X1600 works in 10.4.11 and 10.5.2, *not* in 10.5.3
unfortunately...
I'm still writing up a Radar with some repro code kindly provided by a
colleague of mine -- I'll post the number here for reference once it's
ready.
Sigh,
Dario
On 30 May 2008, at 22:47, Ben Supnik wrote:
Hi Y'all,
I'm getting reports from users that one of our textures (a shadow) is
incorrect after they got the 10.5.3 update. Has anyone else seen
this? I don't have that particular combination of hw/software in my
lab. :-(