I was recently investigating a gl-error that appears some times while
using virgl, where I got a pretty nonsensical GL-error:
Mesa: User error: GL_INVALID_VALUE in glGetnTexImageARB(depth = 0)
Now, the reason why this is nonsensical, is that glGetnTexImage doesn't
have a 'depth'-argument, this value is generated internally in mesa.
What happens is that virgl ends up asking for the teximage for a
non-existen mipmap-level. Whoops, that shouldn't be done, but while
we're in this area, let's clean this up.
My first fix was to revert 42891dbaa12 ("gettextsubimage: verify zoffset
and depth are correct"). But that would render that commit moot, but it
actually does fix something. So I'd rather not; it's a good change, but
this logic should only apply to glGetTextureSubImage.
So I decided to give glGetTextureSubImage its own error checking.
While working on this, I also realized that the error introduced in
b37b35a5d26 ("getteximage: assume texture image is empty for non defined
levels") was flawed; it shouldn't apply in the case where *all* of
width, xoffset and the texture's width are zero. And since we end up
checking explicitly for these conditions later, this should simply be a
no-op as per the spec. And we already have some support-code for this.
No regressions found in piglit.
Erik Faye-Lund (6):
mesa/main: remove ARB suffix from glGetnTexImage
mesa/main: remove bogus error for zero-sized images
mesa/main: factor out tex-image error-checking
mesa/main: factor out common error-checking
mesa/main: check cube-completeness in common code
mesa/main: fix incorrect detph-error
src/mapi/glapi/gen/ARB_robustness.xml | 2 +-
src/mapi/glapi/gen/gl_API.xml | 9 +
src/mesa/main/texgetimage.c | 273 ++++++++++++++++----------
src/mesa/main/texgetimage.h | 4 +-
4 files changed, 179 insertions(+), 109 deletions(-)
--
2.19.1
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
