[Mesa-dev] fixes for 8.0

[Andreas Boll]

Hi list,

Actually I wanted to wait for my fdo account and then push a branch at
people.fdo with
the candidates for 8.0. But in the meantime I'm sending you an update
with some issues to solve.

I have cherry-picked almost all of the outstanding candidates to mesa 8.0
The following commits have caused conflicts:

- error: could not apply 19bd593... i965: Add support for
GL_SKIP_DECODE_EXT on other SRGB formats.
- error: could not apply c0f6010... i965/fs: Don't clobber sampler
message MRFs with subexpressions.
- error: could not apply 9b4053c... i965: Drop the confusing saturate
argument to math instruction setup.
- error: could not apply e119f98... i965/vs: Add support for copying
user edge flags.
- error: could not apply 4e087de... intel: Reserve enough space to
finish occlusion queries on Gen6.
- error: could not apply 6928bea... i965/fs: Initialize
output_components[] by filling it with zeros.

And the following commit does not compile on the 8.0 branch, since the
functions _mesa_is_desktop_gl(),
_mesa_is_gles() and _mesa_is_gles3() do not exist in the 8.0 branch.
It has to be manually backported.

- 3f7c8364cf1d763f04413aecfac66df5534aeb34 mesa: Filter glBindTexture
targets based on supported features.

And finally this is the git-shortlog of the successfully cherry-picked commits:

Alex Deucher (3):
      r600g: 8.0.x support for Trinity
      r600g: add new Sumo, Palm, BTC pci ids
      r600g: add additional evergreen pci ids

Brian Paul (11):
      mesa: loosen small matrix determinant check
      radeon: set swrast_renderbuffer::ColorType field when mapping
renderbuffers
      xlib: add X error handler around XGetImage() call
      svga: fix invalid memory reference in needs_to_create_zero()
      meta: fix glDrawPixels fallback test, stencil drawing
      st/mesa: fix glCopyTexSubImage crash
      gallivm: fix crash in lp_sampler_static_state()
      st/mesa: fix renderbuffer validation bug
      softpipe: fix softpipe_delete_fs_state() failed assertion
      mesa: raise GL_INVALID_OPERATION in glGenerateMipmap for missing
base image
      st/mesa: s/CALLOC/calloc/ to fix allocation bug

Chad Versace (2):
      swrast: Fix implicit declaration warnings
      i830: Fix stack corruption

Christoph Bumiller (1):
      st/mesa: call update_renderbuffer_surface for sRGB renderbuffers, too

Eric Anholt (2):
      i965/vs: Convert EdgeFlagPointer values appropriately for the VS on gen4.
      i965: Fix accumulator_contains() test to also reject swizzles of the dst.

Ian Romanick (3):
      dri2: Fix bug in attribute handling for non-desktop OpenGL contexts
      mesa: Generate an error when glCopyTexImage border is invalid
      mesa/es: Validate glTexImage border in Mesa code rather than the
ES wrapper

Jordan Justen (1):
      intel: move error on create context to proper path

José Fonseca (1):
      mesa: disable MSVC global optimization in pack.c

Kenneth Graunke (4):
      mesa: Prevent repeated glDeleteShader() from blowing away our refcounts.
      i965/vs: Don't clobber sampler message MRFs with subexpressions.
      intel: Move finish_batch() call before MI_BATCH_BUFFER_END and padding.
      i965/fs: Don't use brw->fragment_program in calculate_urb_setup().

Marek Olšák (1):
      gallium/u_blit: set dst format from pipe_resource, not pipe_surface

Niels Ole Salscheider (1):
      st/mesa: index can be negative in the PROGRAM_CONSTANT case

Stéphane Marchesin (1):
      glsl/linker: Avoid buffer over-run in
parcel_out_uniform_storage::visit_field


I have done a piglit (1d05317) quick run with r600g on my rv770
without any regressions.


Andreas