radeonsi: clear DB_RENDER_OVERRIDE

Vulkan doesn't set these fields even though it doesn't use HiS.
HiS is disabled by programming DB_SRESULTS_COMPARE_STATEn to 0.

Reviewed-by: Nicolai Hähnle <nicolai.haehnle@amd.com>

glsl: Disable textureOffset(sampler2DArrayShadow, ...) in GLSL ES.

This has apparently never existed in GLSL ES.

Fixes dEQP-GLES3.functional.shaders.texture_functions.invalid
.textureoffset_sampler2darrayshadow_vec4_ivec2_vertex and
.textureoffset_sampler2darrayshadow_vec4_ivec2_fragment

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98244
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>

st/nine: Fix multisample limit check

Fixes regression introduced by
b5603056872708fdd82f1224854097805a01d4c0

The regression prevents some apps to start.

Signed-off-by: Axel Davy <axel.davy@ens.fr>

vc4: Fix fast clear color packing for 565.

Piglit didn't manage to cover this because fbo-clear-formats uses
scissors, so we don't get fast clearing.

state_tracker: Fix check for scissor enabled when < 0.

DEQP's clear tests like to give us x + w < 0 or y + h < 0.  Since we
were comparing to an unsigned, it would get promoted to unsigned and come
out as bignum >= width or height and we would clear the whole fb instead
of none of the fb.

Fixes 10 tests under deqp-gles2/functional/color_clear.

Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>
Acked-by: Edward O'Callaghan <funfunctor@folklore1984.net>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>

egl/surfaceless: Fix comparison between pointer and integer

Fixes GCC warning:
  drivers/dri2/platform_surfaceless.c:196:18: warning: comparison
      between pointer and integer

Fixes: 4b8a55809eb ("egl/surfaceless: tweak surfaceless_add_configs_for_visuals()")
Reviewed-by: Emil Velikov <emil.l.velikov@gmail.com>

egl/surfaceless: use correct index when accesing the visual

i is used for the driver_configs, while j is for the visuals.

Fixes: 4b8a55809eb ("egl/surfaceless: tweak surfaceless_add_configs_for_visuals()")
Reported-by: Chad Versace <chadversary@chromium.org>
Tested-by: Chad Versace <chadversary@chromium.org>
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

radv/winsys: Fail early on overgrown cs.

When !use_ib_bos, we can't easily chain ibs one to another. If the
required cs size grows over 1Mi - 8 dwords just fail the cs so that we
won't assert-fail in radv_amdgpu_winsys_cs_submit later on.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>

glsl: Drop the ES requirement that VS outputs must be flat qualified.

Several conformance tests violate this requirement:

ES31-CTS.core.tessellation_shader.max_patch_vertices
ES31-CTS.core.tessellation_shader.tessellation_control_to_tessellation_evaluation.data_pass_through

I submitted a merge request to fix the conformance tests, but Khronos
opted to drop this GLSL ES specific requirement in favor of making flat
qualification of VS outputs optional, matching modern desktop GL.

Note that there were 7 Piglit tests which enforce this rule:
tests/spec/glsl-es-3.00/compiler/interpolation/qualifiers/*nonflat*
but these were deleted in Piglit commit acc0a2fabbd714bc704c16f1675e7c0.

Bugzilla: https://cvs.khronos.org/bugzilla/show_bug.cgi?id=15465#c7
Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Timothy Arceri <timothy.arceri@collabora.com>

intel/genxml: Make some PIPE_CONTROL fields booleans

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel/genxml: Make "Predication enable" a boolean

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel/genxml; Make "Use Global GTT a boolean

We also remove the redundant zero defaults since everything without an
explicit default gets zeroed automatically.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel/genxml; Make "Tiled Surface" a boolean

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel/genxml: Make "SO Buffer Enable" fields boolean

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel/genxml: Make "Stencil Buffer Enable" a boolean

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel/genxml: Make a couple of STREAMOUT fields booleans

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel/genxml: Make "Include Vertex Handles" and "Include Primitive ID" booleans

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel/genxml: Make "Vector Mask Enable" a boolean

We also get rid of the "(VME)" a few places

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

intel/genxml: Make "Single Program Flow" a boolean

We also get rid of the "(SPF)" a few places.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Kenneth Graunke <kenneth@whitecape.org>

nv50/ir: constant fold OP_SPLIT

Split the source immediate value into new values and move them into the
original defs set by the split. Since we can only have up to 64-bit
immediates, this is largely beneficial for F64 (and, in the future, U64)
operations.

Signed-off-by: Tobias Klausmann <tobias.johannes.klausmann@mni.thm.de>
[imirkin: always use U32, set newi for foldCount tracking]
Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>

i965: Enable OpenGL 4.5.

Everything is in place.  There are still conformance issues to sort out,
but we may as well turn it on in master.

Signed-off-by: Kenneth Graunke <kenneth@whitecape.org>
Reviewed-by: Ian Romanick <ian.d.romanick@intel.com>
Acked-by: Jason Ekstrand <jason@jlekstrand.net>

anv/pipeline: Remove a meta hack from emit_ds_state

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>

anv/image: Create views directly in VkCreate*View

Without meta, we no longer need the _init helpers and the ability to back
an image view with surface states allocated out of the command buffer.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>

anv/image: Get rid of the usage hacks for meta

Now that meta is gone and we're using blorp, we don't need all of the usage
hacks.  Instead, the usage provided by the app is exactly the usage that we
want because the app is the only thing creating image views.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>

anv: Move Create*Pipelines into genX_cmd_buffer.c

Now that we don't have meta, we have no need for a gen-agnostic pipeline
create path.  We can, instead, just generate one Create*Pipelines function
per gen and be done with it.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>

anv/pipeline: Remove support for direct-from-nir shaders

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>

anv: Make entrypoint resolution take a gen_device_info

In order for things such as the ANV_CALL and the ifuncs to work, we used to
have a singleton gen_device_info structure that got assigned the first time
you create a device.  Given that the driver will never be used
simultaneously on two different generations of hardware, this was fairly
safe to do.  However, it has caused a few hickups and isn't, in general, a
good plan.  Now that the two primary reasons for this singleton are gone,
we can get rid of it and make things quite a bit safer.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>

anv: Get rid of the ANV_CALL macro

This macro was needed by meta in order to make gen-specific calls from
gen-agnostic code.  Now that we don't have meta, the remaining two uses are
fairly trivial to get rid of.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>

anv: Get rid of graphics_pipeline_create_info_extra

Now that we no longer have meta, all pipelines get created via the normal
Vulkan pipeline creation mechanics.  There is no more need for this bit of
extra magic data that we've been passing around.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

anv: Get rid of meta

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

anv: Use blorp for subpass clears

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

anv: Use blorp for ClearAttachments

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

anv/hiz: Perform HiZ resolves for all partial renders

If we don't, we can end up with corruption in the portion of the depth
buffer that lies outside the render area when we do a HiZ resolve at the
end.  The only reason we weren't seeing this before was that all of the
meta-based clears such as VkCmdClearDepthStencilImage were internally using
HiZ so the HiZ buffer never truly got out-of-sync.  If the CTS ever tested
a depth upload (which doesn't care about HiZ) and then a partial render we
would have seen problems.  Soon, we will be using blorp to do depth clears
and it won't bother with HiZ so we would get CTS regressions without this.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Nanley Chery <nanley.g.chery@intel.com>

anv: Use blorp for ClearDepthStencilImage

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

anv/image: Add an isl_view to anv_image_view

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

anv/image: Rework our handling of 3-D image array ranges

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

anv/blorp: Don't hand-roll flush_pipeline_select_3d

When I initially brought up Vulkan blorp, I completely missed that this
was already factored out.  There's no good reason for us to hand-roll it.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/blorp: Add a flag to make blorp not re-emit dept/stencil buffers

In Vulkan, we want to be able to use blorp to perform clears inside of a
render pass.  If blorp stomps the depth/stencil buffers packets then we'll
have to re-emit them.  This gets tricky when secondary command buffers get
involved.  Instead, we'll simply guarantee that the depth and stencil
buffers we pass to blorp (if any) match those already set in the hardware.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/blorp: Add an entrypoint for clearing depth and stencil

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/blorp: Emit a NULL render target for depth/stencil-only operations

This never mattered before because the only time we used blorp
depth/stencil only was to do HiZ operations on gen6-7.  It may have worked
in that case (and maybe it didn't) but slow depth clears actually do depth
rendering so they need a valid render target.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/blorp: Allow for running without a PS on gen8+

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/blorp: Add an "enabled" bit to surface_info

This gives a slightly smarter way to check whether or not a particular
surface exists than looking at the address.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/blorp: Emit more complete DEPTH_STENCIL state

This should now set the pipeline up properly for doing depth and/or stencil
clears by plumbing through depth/stencil test values.  We are now also
emitting color calculator state for blorp operations without an actual
shader because that is where the stencil reference value goes pre-SKL.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/blorp: Unify the DEPTH_STENCIL emit code across gens

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/blorp: Simplify depth/stencil config

The newly reworked depth/stencil config code can properly handle having
depth, stencil, both, or neither.  We no longer need to predicate it on
having depth or stencil.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/blorp: Set QPitch for depth and HiZ on gen8+

intel/blorp: Add support for binding an actual stencil buffer

While we're here, we also make depth without HiZ work.

v2:
 - Use the correct surface type for 1-D on SKL+
 - Set QPitch on BDW+

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/blorp: Move CLEAR_PARAMS setup into emit_depth_stencil_config

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

intel/genxml: Add a uint MOCS field to 3DSTATE_STENCIL_BUFFER

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>

intel/blorp: Make the Z component of the primitive adjustable

We want to be able to start doing slow depth clears with blorp.  This
allows us to adjust the depth we're clearing to.

Signed-off-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Topi Pohjolainen <topi.pohjolainen@intel.com>

i915: workaround multiple intelFenceExtension definitions

Due to conflicting symbol names (between i915 and i965) in the
megadriver, we use a set of defines in i915/intel_screen.h.

With a recent commit we've introduced a symbol intelFenceExtension which
has different implementation for each driver, yet we forgot to add the
define.

Fixes: d11515ff1b3 ("i915/sync: Implement DRI2_Fence extension")
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=98264
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>

docs/specs: Update allocated EGL enum values

Document the EGL enum ranges for Mesa and those values allocated by the
following extensions:

    EGL_MESA_drm_image
    EGL_MESA_platform_gbm
    EGL_MESA_platform_surfaceless
    EGL_WL_bind_wayland_display

Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

doc/specs: Reference the Khronos registry XML

Years ago Khronos replaced the registry's spec files with newfangled XML
files.  Update the reference in doc/specs/enum.txt accordingly.

Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

egl: Move old EGL_MESA_screen_surface spec

It was the lone file in src/egl/docs. Move it to where the other specs
live, in $MESA_TOP/docs/specs.

Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

egl: Implement EGL_MESA_platform_surfaceless

Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

egl: Don't advertise unsupported platform extensions

Mesa's set of supported platform extensions depends on the autoconf
option --with-egl-platforms=foo,bar,baz. If --with-egl-platforms lacks
foo, then eglGetPlatformDisplay(EGL_PLATFORM_FOO, ...) unconditonally
fails.

So, if --with-egl-platforms lacks foo, then remove
EGL_VENDOR_platform_foo from the EGL client extension string.

Cc: mesa-stable@lists.freedesktop.org
Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

docs: Add EGL_MESA_platform_surfaceless.txt (v2)

v2:
    - Assign enum values.
    - Define interactions with EGL_EXT_platform_base and EGL 1.4.

Reviewed-by: Emil Velikov <emil.velikov@collabora.com>

i965: Sort some extension names

Trivial.

Signed-off-by: Ian Romanick <ian.d.romanick@intel.com>
Reviewed-by: Eduardo Lima Mitev <elima@igalia.com>

scons: Fix the Python dependency scanner.

modulefinder wasn't searching for dependencies in the script dir.

It's not capable of detecting the sys.path manipulations scripts do
internally neither.

This change fixes the first issue, and hacks around the second.

Honestly, I've come to the conclusion that automatic Python dependency it will always be
too brittle.   I think we should start manually typing the dependencies
like we do in automake.  At very least it will enable any person to
eyeball and spot/fix missing dependencies, without dig into SCons internals.

pipe_loader_sw: Don't invoke Unix close() on Windows.

Trivial.

Revert "egl/dri2: rework dri2_make_current code flow"

This reverts commit 675719817e7bf7c5b9da22c02252aca77a41338d.

i915: store reference to the context within struct intel_fence (v2)

Porting of the corresponding patch for i965.

Here follows the original commit message by Tomasz Figa:

"As the spec allows for {server,client}_wait_sync to be called without
currently bound context, while our implementation requires context
pointer.

v2: Add a mutex and acquire it for the duration of
    brw_fence_client_wait() and brw_fence_is_completed() as suggested
    by Chad."

NOTE: in i915 all references to 'brw' are replaced by 'intel'

Marshmallow-x86 boots ok with the following results of Android CTS.

Android CTS 6.0_r7 build:2906653
Session     Pass       Fail     Not Executed
0(EGL)      1410       24       0
1(GLES2)    13832      82       0

I get the same results as per i965GM.

[Emil Velikov: Include Mauro's test results]
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>

i915/sync: Implement DRI2_Fence extension

Here is the porting of corresponding patch for i965,
i.e. commit c636284 i965/sync: Implement DRI2_Fence extension

Here follows part of original commit message by Chad Versace:

"This enables EGL_KHR_fence_sync and EGL_KHR_wait_sync."

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>

i915/sync: Replace prefix 'intel_sync' -> 'intel_gl_sync'

This is the porting of corresponding patch for i965,
i.e. commit 2516d83 i965/sync: Replace prefix 'intel_sync' -> 'intel_gl_sync'

The only difference compared to i965 one is that intel_check_sync() was renamed
to intel_gl_check_sync() here, as it is more appropriate.

Here follows original commit message by Chad Versace:

"I'm about to implement DRI2_Fenc in intel_syncobj.c.  To prevent
madness, we need to prefix functions for GL_ARB_sync with 'gl' and
functions for DRI2_Fence with 'dri'. Otherwise, the file will become
a jumble of similiarly named functions.

For example:
    old-name:      intel_client_wait_sync()
    new-name:      intel_gl_client_wait_sync()
    soon-to-come:  intel_dri_client_wait_sync()

I wrote this renaming commit separately from the commit that implements
DRI2_Fence because I wanted the latter diff to be reviewable."

[Emil Velikov: rename the outstanding intel_sync instances]
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>

egl/drm: set eglError and provide an error message on failure

v2: Remove gratuitous newline/semicolon (Eric)

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/x11: attribute for dri2_add_config failure

... in dri2_x11_add_configs_for_visuals().

Currently the latter does not consider that, thus in such cases it adds
"empty" configs in the list.

Properly account for things and as we do that we can reuse count,
instead of calling _eglGetArraySize to determine if we've added any
configs.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/wayland: introduce dri2_wl_add_configs_for_visuals() helper

Analogous to previous commits - with an extra bonus.

Current code, apart from not attributing the lack of 'per visual'
and overall configs also overwrites the newly added config.

Namely if the dpy supports two or more of the supported formats
(XRGB8888, ARGB8888 and RGB565) earlier configs will be overwritten
and the the final one will be stored, since the we use the same index
for all three in our dri2_add_config call.

v2: Use correct comparison in loop conditional (Eric)
    Use valid C initializer (Gurchetan)

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/surfaceless: tweak surfaceless_add_configs_for_visuals()

Analogous to previous commit.

v2: Use correct comparison in loop conditional (Eric)
    Use valid C initializer (Gurchetan)

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/android: tweak droid_add_configs_for_visuals()

Iterate over the driver_configs first in order to cut down the number of
getConfigAttrib() calls by a factor of 5.

While we're here, also drop the sentinel of the visuals array. We
already know its size so we can use that and save a few bytes.

v2: Use correct comparison in loop conditional (Eric)
    Use valid C initializer (Gurchetan)

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/drm: introduce drm_add_configs_for_visuals() helper

Factor out and rework the existing code so that it prints a debug
message if we have zero configs for any visual.

As a nice side effect we now provide a correct (sequential ID) when
creating a config (via dri2_add_config).

v2: Use correct comparison in loop conditional (Eric)
    Use valid C initializer (Gurchetan)

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/surfaceless: print out a message on zero configs for given format

Currently we print a debug message if the total configs is non-zero only
to do the same (at an error level) as we return from the function.

Rework the message to print if we're missing a config for the given
format.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/dri2: set WL_bind_wayland_display in a consistent way

Introduce a helper and use it throughout the platform code. This allows
us to reduce the amount of ifdef(s) and (potentially) use
kms_swrast_dri.so for !drm platforms (namely wayland and x11).

Note: in the future as other platforms (android, surfaceless) support
the extension they can reuse the helper.

v2: Rebase, check for device_name.
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Daniel Stone <daniels@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/android: remove duplicate KHR_image_base set

The core egl/dri2 already sets the extension bit _only_ when possible -
which in Android's case is always.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>

loader/dri3: constify the loader_dri3_vtable

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/dri2: micro optimise dri2_bind_extensions()

Do not loop over all matches if we've already found one.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/dri2: annotate dri2_extension_match instances as const data

v2: Rebase.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/dri2: use dri2_bind_extensions to manage the optional extensions

v2: dri2_bind_extensions() now takes optional as an argument.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

gbm: rename gbm_dri_device::{,loader_}extensions

To align with the name used in the EGL and GLX loaders.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/dri2: add support for optional extensions in dri2_bind_extensions()

Will allow us to reuse the function for optional extensions and fold a
bit of code.

v2: Make dri2_bind_extensions::optional flag an argument to
dri2_bind_extensions (Kristian).

Cc: Rob Clark <robdclark@gmail.com>
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>

egl/dri2: coding style cleanup

Consistently indent with space rather than a mix of tab and
spaces.

v2: Keep the structs properly aligned (Eric).

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>

egl/x11: don't crash if dri2_dpy->conn is NULL

The dri3 version of commits 60e9c35b3a0 and 6de9a03bed4.

While using xcb_connect() guarantees that we always get a non NULL
return value, XGetXCBConnection() does/can not.

CC: <mesa-stable@lists.freedesktop.org>
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/dri2: rework dri2_egl_display::extensions storage

Remove the error prone fixed size array.
While we're here also rename to loader_extensions like in the GLX code.

v2: Rebase. Keep image_loader_extension within the wayland_drm
dri2_loader_extensions list.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/dri2: remove unused dri2_egl_display::{dri2,swrast}_loader_extension

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/x11: don't populate dri2_dpy->swrast_loader_extension

Analogous to earlier commits.

Note: the actual version of the extension is 1, since it does not
implement .putImage2.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/wayland: don't populate dri2_dpy->swrast_loader_extension

Similar to the dri2 one - the extension stored in struct
dri2_egl_display is unused.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/x11: don't populate dri2_dpy->dri2_loader_extension

Analogous to the earlier android and wayland patches. As we're here we
can drop exposing the old version of the extension.

Any dri loader/driver interface use lower bound checking thus exposing
dri2 loader v3 to a v2 capable driver is perfectly normal.

v2: Preserve compat with dri2_minor < 1. The driver does not know if
there is a protocol to manage getBuffersWithFormat(). It's up-to the
loader to expose the vfunc if there is one. (Kristian)

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/wayland: don't populate dri2_dpy->dri2_loader_extension

Analogous to the earlier android patch.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/surfaceless: trivial coding style fixes

Remove a few gratious blank lines and use the correct level of
indentation.

Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/surfaceless: don't check the mask(s) prior to calling dri2_add_config

The latter already does it for us.

As we're here annotate the masks as const and use unsigned for the
index(es).

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>
Reviewed-by: Eric Engestrom <eric@engestrom.ch>

egl/surfaceless: remove unused dri2_loader_extension implementation

Earlier commit introduced support for image_loader and left the
dri2_loader code dangling/unused. Let's remove it.

Fixes: 63c5d5c6c46 ("Added pbuffer hooks for surfaceless platform")
Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Gurchetan Singh <gurchetansingh@chromium.org>

egl/android: don't populate dri2_dpy->dri2_loader_extension

The extension stored in struct dri2_egl_display isn't used, thus we can
create a static const instance of the extension and point extensions[]
to it.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>

egl/dri2: rework dri2_make_current code flow

Fold duplicate conditional blocks and add a few extra comments ;-)

v2: Bring back the explicit "unbind" logic (Eric), remove NULL derefs.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>

egl/dri2: drop NULL checks prior to dri2_destroy_surface

The function already have the respective check within.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>

egl/dri2: call static functions directly, not via _EGLDriver::API

The indirection is meant to be used by the core EGL implementation in
main. Not in the drivers themselves.

Move the dri2_destroy_surface definition to avoid forward declaration of
the static function.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

egl/dri2: use dri2_egl_display inline wrapper where possible

This way the only places that reference DriverData are the ones that
manipulate it.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

egl/dri2: bail out on NULL dpy in dri2_display_release()

Currently all callers are careful enough not to do that, yet that will
not be the case in the future.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

egl/dri2: move surface refcounting out of the platform code

All the platforms are duplicating what should be a driver/dri2 thing -
refcounting. Just fold it accordingly.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

egl/dri2: coding style fix

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

egl/dri2: non-shared glapi cleanups

For a while now we require shared glapi for EGL, thus we can drop a
few bits from the olden days. Namely - dlopen(NULL...) is not possible,
error out at build stage if so and drop the guard around dlclose().

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>

egl/dri2: glFlush is not optional, treat it as such

The documentation is clear - one must glFlush the old context on
eglMakeCurrent. Thus keeping it optional is not something we should be
doing. Furthermore if we cannot get the entry point we're likely having
a broken setup/stack.

Signed-off-by: Emil Velikov <emil.velikov@collabora.com>
Reviewed-by: Eric Engestrom <eric.engestrom@imgtec.com>