Mesa 17.2.1 Release Notes / September 17, 2017

Mesa 17.2.1 is a bug fix release which fixes bugs found since the 17.2.0 release.

Mesa 17.2.1 implements the OpenGL 4.5 API, but the version reported by glGetString(GL_VERSION) or glGetIntegerv(GL_MAJOR_VERSION) / glGetIntegerv(GL_MINOR_VERSION) depends on the particular driver being used. Some drivers don't support all the features required in OpenGL 4.5. OpenGL 4.5 is only available if requested at context creation because compatibility contexts are not supported.

SHA256 checksums

c902d8dc2540195bc570d88af1a8fd8a1774373660a27bb1d539551f46824bc1  mesa-17.2.1.tar.gz
77385d17827cff24a3bae134342234f2efe7f7f990e778109682571dbbc9ba1e  mesa-17.2.1.tar.xz

New features

None

Bug fixes

• Bug 100613 - Regression in Mesa 17 on s390x (zSystems)
• Bug 101709 - [llvmpipe] piglit gl-1.0-scissor-offscreen regression
• Bug 102454 - glibc 2.26 doesn't provide anymore xlocale.h
• Bug 102467 - src/mesa/state_tracker/st_cb_readpixels.c:178]: (warning) Redundant assignment
• Bug 102502 - [bisected] Kodi crashes since commit 707d2e8b - gallium: fold u_trim_pipe_prim call from st/mesa to drivers

Changes

Bas Nieuwenhuizen (4):

• radv: Actually set the cmd_buffer usage_flags.
• radv: Fix vkCopyImage with both depth and stencil aspects.
• radv: Disable multilayer & multilevel DCC.
• radv: Don't allocate CMASK for linear images.

Ben Crocker (1):

• llvmpipe: lp_build_gather_elem_vec BE fix for 3x16 load

Brian Paul (1):

• llvmpipe: initialize llvmpipe->dirty with LP_NEW_SCISSOR

Charmaine Lee (1):

• vbo: fix offset in minmax cache key

Dave Airlie (12):

• radv: disable 1d/2d linear optimisation on gfx9.
• radv/gfx9: set descriptor up for base_mip to level range.
• Revert "radv: disable support for VEGA for now."
• radv/winsys: use amdgpu_bo_va_op_raw.
• radv/gfx9: allocate events from uncached VA space
• radv: use simpler indirect packet 3 if possible.
• radv: don't use iview for meta image width/height.
• radv: handle GFX9 1D textures
• radv/gfx9: set mip0-depth correctly for 2d arrays/3d images
• radv/ac: bump params array for image atomic comp swap
• radv/gfx9: fix image resource handling.
• radv/winsys: fix flags vs va_flags thinko.

Emil Velikov (7):

• docs: add sha256 checksums for 17.2.0
• cherry-ignore: add getCapability patches
• cherry-ignore: ignore gfx9 tile swizzle fix
• cherry-ignore: add execution_type() fix to the list
• cherry-ignore: add EGL+gbm swast patches
• egl/x11/dri3: adding missing __DRI_BACKGROUND_CALLABLE extension
• Update version to 17.2.1

Eric Engestrom (3):

• util: improve compiler guard
• mesa/st: remove unwanted backup file
• docs/egl: remove reference to EGL_DRIVERS_PATH

Grazvydas Ignotas (1):

• radv: don't assert on empty hash table

Jason Ekstrand (2):

• anv/formats: Nicely handle unknown VkFormat enums
• spirv: Add support for the HelperInvocation builtin

Karol Herbst (1):

• nvc0: write 0 to pipeline_statistics.cs_invocations

Kenneth Graunke (2):

• i965: Fix crash in fallback GTT mapping.
• i965: Set "Subslice Hashing Mode" to 16x16 on Apollolake.

Marek Olšák (1):

• st/mesa: skip draw calls with pipe_draw_info::count == 0

Michael Olbrich (1):

• egl/dri2: only destroy created objects

Nicolai Hähnle (1):

• radeonsi: apply a mask to gl_SampleMaskIn in the PS prolog

Nicolai Hähnle (4):

• radeonsi/gfx9: always flush DB metadata on framebuffer changes
• st/glsl_to_tgsi: only the first (inner-most) array reference can be a 2D index
• ac/surface: match Z and stencil tile config
• glsl: fix glsl_struct_field size calculations for shader cache

Ray Strode (1):

• gallivm: correct channel shift logic on big endian

Rob Clark (1):

• freedreno: skip batch-cache for compute shaders

Roland Scheidegger (1):

• st/mesa: fix view template initialization in try_pbo_readpixels

Samuel Pitoiset (1):

• radeonsi: update dirty_level_mask before dispatching

Timothy Arceri (9):

• glsl: allow NULL to be passed to encode_type_to_blob()
• glsl: stop adding pointers from gl_shader_variable to the cache
• glsl: stop adding pointers from glsl_struct_field to the cache
• glsl: add has_uniform_storage() helper to shader cache
• glsl: don't write uniform storage offset if there isn't one
• glsl: always write a name/label string to the cache
• compiler: move pointers to the start of shader_info
• glsl: stop adding pointers from shader_info to the cache
• glsl: stop adding pointers from bindless structs to the cache