If the currently bound FBO isn't yet validated it's possible for
rgbMode to be zero so we'll lose the texture enable.
This could fix some FBO rendering glitches, but I don't know of
any specific instances.
Mostly fixes progs/demos/lodbias when MESA_TEX_PROG=1. But the LOD still
seems off by -1 or so.
May be an issue with the params passed to _swrast_compute_lambda()
In get_array_bounds we were previously defining a user buffer sized as
(nr_vertices * stride). The trouble is that if the vertex data
occupies less than stride bytes, the extra tailing (stride - size)
bytes may extend outside the memory actually allocated by the app and
caused a segfault.
To fix this, define a the buffer bounds to be:
ptr .. ptr + (nr-1)*stride + element_size
The warnings introduced in 1f309c40b8
would pour out generously from some applications. This patch adds a
"warn once" wrapper macro, heavily inspired by
src/mesa/drivers/dri/r600/radeon_debug.h
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
This happens to rendering with textures with a border, which had resulted
in a segfault on dereferencing the irb.
(cherry-picked from commit 8bba183b9e)
Since commit 2921a2555d ('intel: Deassociated
drawables from private context struct in intelUnbindContext'),
intel->driDrawable may be NULL in intel_flush().
Previously srandom and random were used. This cause the global random
number generator state to be modified. This caused problems for
applications that called srandom before calling into GLX. By using
local state the global state is left unmodified.
This should fix bug #23774.
The following example caused an incorrect GL_OUT_OF_MEMORY error to be
raised in glTexSubImage2D:
glTexImage2D(level=0, width=32, height=32, pixels=NULL);
glTexImage2D(level=0, width=64, height=64, pixels=NULL);
glTexSubImage2D(level=0, pixels!=NULL);
The second glTexImage2D() call needs to cause the first image to be
deallocated then reallocated at the new size. This was not happening
because we were testing for pixels==NULL too early.
The generic DRI infrastructure makes sure that __DRIcontextRec::driDrawablePriv
and __DRIcontextRec::driReadablePriv are set to NULL after unbinding a
context. However, the intel_context structure keeps cached copies of
these pointers. If these cached pointers are not NULLed and the
drawable is actually destroyed after unbinding the context (typically
by way of glXDestroyWindow), freed memory will be dereferenced in
intelDestroyContext.
This should fix bug #23418.
When we concatenate shaders to do our form of poor-man linking, if there's
multiple #version directives, preprocessing fails. This change disables
the extra #version directives by changing the first two chars to //.
This should help with some Wine issues such as bug 23946.
This fixes a bug seen when doing a glDrawPixels(GL_STENCIL_INDEX) right
after a glClear(). The check-for-flush test was failing because we
didn't set the dirty_render_cache flag in softpipe_clear(). So we saw
stale data when we mapped the stencil buffer.
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Hopefully this will be one of the last cherry-picks.
(cherry picked from commit ca246dd186)
For some IZ setups, we'd forget to account for the source depth register
being present, so we'd both read the wrong reg, and write output depth to
the wrong reg.
Bug #22603.
(cherry picked from commit f44916414e)
Fixes piglit glsl-vs-if-bool and progs/glsl/twoside, and will likely be
useful for the looping code.
Bug #18992
(cherry picked from commit 78c022acd0)
(cherry picked from commit 63d7a2f53f)
This avoids sending a bad buffer address to the GPU due to programmer error,
and is permitted by the ARB_vbo spec. Note that we still have the opportunity
to dereference past the end of the GPU, because we aren't clipping to a
correct _MaxElement, but that appears to be harder than it should be. This
gets us the 90% solution.
Bug #19911.
(cherry picked from commit d7430d942f)
See comment on Vertex URB Entry Read Length for VS_STATE.
This, combined with the previous three commits, fixes#22945.
(cherry picked from commit e340d4f986)
It appears that sometimes Mesa (and I suppose a VS could as well) emits
a program which references no vertex data, and thus we end up with
nr_enabled == 0 even though some VBs are enabled. We'd end up emitting
VB/VE packet headers of 0xffffffff in that case, leading to GPU hangs.
Bug #22945 (wine with an uncompiled VS)
(cherry picked from commit d1fbfd0f96)
