eglGetProcAddress should return a function pointer that does not
depends on the current bounded context (if any), when the user
calls one of the function pointers returned from eglGetProcAddress
for one of the GLES extension functions, the GLESv1 or GLESv2
version of the extension function shold be called depending on
the current bounded context.
For this we have added a ClientAPI (GLES) extension dispatch
table in the EGL level which points to static functions in libEGL
where each function checks in runtime the current bound context and
calls down to the GLES_CM -or- GLESv2 library function.
See ClientAPIExts.cpp, when new GLES extension functions are added
to GLESv1 or GLESv2 its definition should be added to
ClientAPIExts.in as well.
This fixes the segfault in egl_image conformance test.
Change-Id: I8464d87c2fcbe57d67bd8b891b695a690dec89f3
Handle EGL context and surfaces destrouction using ref counting
pointer (SmartPtr) rather then maintaining a "markForDeletion" flag.
This has required to re-arrange the ThreadInfo structure to be local
to EGL rather then in the GLcommon area so that we will be able to
place a ref-counted pointer to the current context in the TLS.
That fixes the crash during exit of covgl conformance test.
Change-Id: I6f0a2e49c6f21eb75b654cab4346c764715c449b
The was a bug in the calculation of the pixels data size, caused by wrong usage
of pack/unpack alignment.
Pack should be used for glReadPixels, while unpack should be used for glTexImage and such.
Also fixes crash in "Angry Birds" & "Angry Birds Rio"
Change-Id: I099308a57a1ea907d38c59bad056429bc107de98
This is needed to support the rare, but still legal scenario,
when glDrawElemets is used with some index array data is stored in VBO's
while the actual attributes (vertices) data is in immediate mode.
When in immediate mode, we need to process the incodes, in order to know
which vertex data to send, which was impossible without the caching.
This commit introduces a new class GLSharedGroup, which will hold all data
that can be shared by shared contexts (buffers are such data).
This also makes the "Jet Cars Stunts" app work properly.
Change-Id: Ic937080dae461bc8cdf4d10cf37066a6e847f464
In case we're receiving packets larger then the initial size
of the read buffer, we dynamicaly grow (X2 steps) the size of
the buffer. An example for this large buffer is a 4MB texture
in JetCarStunts.
Change-Id: I167caddb731583811c009321d4f8fb2f7eea032c
On every swapBuffers, we make sure to decRef on the prev buffer,
and incRef on the current buffer. This fixes all the anoying
crashes on application exit, and also after the speech windows
on Replica Island
Change-Id: I1981fbf17a24ef6f60cdcf6d7ec3a5e7eef5377f
This change modifies the function declarations under
developement/tools/emulator/opengl/host/include/render_api.h
to make them callable from C.
This is preparation work for integrating the rendering library
into the emulator. The plan is to dlopen() the library dynamically
and using non-mangled function names makes using dlsym() both
easier and more portable.
Change-Id: I34656ea4618dbb989fb6ff78df43e9bfb38a7799
Adds GLSL built in constants defined in GLSL ES but not in GLSL.
Makes biConstant conformance test to pass on Mac (where ES2_compatability is not
present).
Also fixed the GLSL parser to insert our additions to the shader
only after the "#version" token if exist in the original shader.
Also renamed env var GOOGLE_GLES_FORCE_GLSL_VERSION to
GOOGLE_GLES_DEFAULT_GLSL_VERSION to better match its functionality since
we use it only if no specific version exist in the shader code.
Change-Id: If9485da16da2808245943a7295f8fb22a6f631aa
Both the ATI and nVidia OpenGL drivers return the wrong answer on
glGetIntegerv(GL_ALPHA_TEST_REF). So Implement this functionality
in our code.
Change-Id: I022b1475f7edb9f1d0f48ebcaf9f6e9f5c4f344e
enable workarounds by default
added environemnt variable GOOGLE_GLSL_FORCE_GLES_VERSION
to the shader parser
Change-Id: If1dbc10722436659295c40a40bd630a72e575799
nvidia may return GL_VALIDATE_STATUS = GL_TRUE even when the program does not link
this change checks for GL_LINK_STATUS when GL_VALIDATE_STATUS is queried
Change-Id: Ia9f5da9c71b8666ce929d8cc2a44639bc947eecb
This is not really a bug, but the nVidia shader compiler is much more
liberal than the GLSL spec, and allows compilation of shaders which
do not comply with the spec. This is an issue for passing the
conformance tests, but more importantly it will allow bad shader written
by the app developer to pass on the emulator and then to fail on the device.
Adding "#version 100" definition to the head of a shader source disables
this liberal behavior. For now this is activated in run time by the "NV_WAR"
environment variable.
Change-Id: I1c322d4221d313d3ee70592bc15ea0e340853990
When calling RenderbufferStorage with internalformat=GL_RGB565
we will create one with internal format GL_RGB
the same for glTexImage2D
Change-Id: I1e464dd36142f586beef6bf895456c7aab1b8237
support different binding to GL_TEXTURE_2D and GL_TEXTURE_CUBE_MAP
targets
create 2 default texture object, to be bound when BindTexture( ... ,0)
is called.
in order to do this, we must ensure their names are internal only, so the user can't
use these names by accident.
the object name manager local name is now 64 bit to allow generation of internal names
Change-Id: I29d000a67042051706e13dc3ef3aa34a1a24f8ca
Do not allow glLinkProgram to succeed if only one shader type has been
attached to the program, GL allows this, but GLES2 does not.
added state tracking for program objects to check which shaders are
attached to the program object.
add a way to define object data type being held for each object in objectNameManager
ProgramData and ShaderParser are both valid objectData for objects in SHADER namespace
we need a way to determine object type to generate correct errors
Change-Id: Ic232549df0bb6daf6ec528cb039482cd68e896bb
This fixes the Windows build of the opengl emulation code.
The official is not broken because you still need to define
BUILD_EMULATOR_OPENGL=true in your environment to enable the
code to be built.
Change-Id: I527f305c8a09df2740a93736123935d480170ffb
Instead of calling the dispatcher directly. This will make sure all Translator's
internals (such as TextureData data) are updated properly.
This also fixes the bug in "Replica Island" - when the green robot was invisible,
because it's TextureData wasn't updated by doCompressedTexImage2D
The pointer to glTexImage2D is passed as a parameter to doCompressedTexImage2D,
otherwise Windows resolve the glTexImage2D symbol in the external libGL rather
than the translator's code. Longer term, we will remove the translator's link
to libGL, at which point this parameter will be removed.
The getGlobalName function returns 0 on error. There are many places
were no error checking is done after calling it to get a shader or
program name. Typically this shouldn't be an issue, since the follwing
OpenGL call using a value of 0 will generated the required GL_INVALID_VALUE,
but this does not always happen.
Change-Id: I8edf2b152978ffd8626f8b4d778cc624f2a3bec3
This fixes a crash caused by covgl (on Nvidia linux driver) when our implementation
of glDrawTexOES calls glDrawArrays with GL_TEXTURE_COORD_ARRAY enabled - but none of texture
units are enabled and no data passed by glTexCoordPointer
Change-Id: Id28b0c35a5a17b139ab196db0e434d2f0a490132
The ARB_ES2_compatibility extension supports several GLES 2.0 enums
and functions on OpenGL 2.0. Seems it exists on nVidia drivers, but
not on ATI's. So implement its functionality in case it doesn't
exist on the host.
Change-Id: Id96f0c8036a7a7d36fca0c85ab5f57bf4b1583aa
in Windows platform we previously defined EGLNativeDisplaytype to be WinDisplay*
( this was defined in include/EGL/eglplatform.h) this was OK as long as we used the
the EGL_DEFAULT_DISPLAY as parameter for eglGetDisplay meaning we were generating
a native display by oureselv and wrapping it, but there was aproblem when we got the
EGLNativeDisplayType from the user (as HDC in Windows) but interpret it as WinDisplay*
this caused a big memory corupption.
so to overcome this problem we now define two types EGLNativeDisplayType to be
HDC and EGLNativeInternalDisplayTypeto be WinDisplay*, and when we get in eglGetDisplay
a parameter different from EGL_DEFAULT_DISPLAY we wrap it with the struct of WinDisplay
and return it to the user.
this change caused a change in all the declerations of EglOsApi.h changing
EGLNativeDisplayType -> EGLNativeInternalDisplayType
Change-Id: I3522c3d507b084c5c211e10ddb6f512d1b90c65e
This commit does few things:
1. Gets doCompressedTexImage2D out of the context and TextureUtils
2. Gets palette textures out of TextureUtils and into PaletteTexture
3. Makes TextureUtils generic for all compressed textures
4. Moves the glGet's dealing with compressed textures from GLEScontext to the impl.
Change-Id: Ibbde6968810a66a4f3488f3887ee3f015fc32c11
Thought the spec doesn't speak about null in data parameter,
it is checked in covgl - so we'll base it on glTexImage2D and allow
null in data.
Change-Id: I9bf5131852bf795acc7ab89bd17b9df90ac99460
This patch allows the emulator_test_renderer program to redirect
input events (mouse, keyboard) to the emulator's console program.
This in order to make testing easier, while we wait for integration
of the display into the emulator program.
For now, this is hard-coded to use localhost port 5554, and only
a few keys are properly translated (Dpad, HOME, BACK, MENU).
Change-Id: Ie6e37c85c291f53de49220a181cb8724f66519ea
This allows the "emulator_test_renderer" program to use
custom framebuffer sizes. You need to define ANDROID_WINDOW_SIZE
in your environment before launching the test program, e.g.:
export ANDROID_WINDOW_SIZE=640x480
Change-Id: I2559c6e75022762697ed4c4559d6c2ba3f400e8f
Support for eglChooseConfig with configs==NULL.
Also added small change to query EGL_LARGEST_PBUFFER attrib of surfaces.
Fixed error print in QemuPipeStream::readFully.
Those changes make sure that covegl completes & passes on the emulator with Translator
on linux in the backend.
Change-Id: I892917ec102ae85d56a16a7cb321e8c12e7083f1
Override glGet* for parameters that request information that have local namespace
convert global names to local names before returning the value
Change-Id: Ie334419b6a5ffe4904d8efe0827a30beaf8b2f6c
fix error generated when glGetVertexAttrib* functions get
index larger or equal to MAX_VERTEX_ATTRIB
INVALID_VALUE should be generated instead of INVALID_ENUM
Change-Id: I0fbacc8fa5b5f14cd302cbe818536cffe5ed3a3c
if points are rendered the built in shader variable gl_PointSize should be active.
added a call to enable VERTEX_PROGRAM_POINT_SIZE to signal opengl to activate this variable
GL_POINT_SPRITE should also be enabled when rendering points
Change-Id: Iba7f62844ee2208ae22700b985aef0229d75fc46
fix behavior of glDeleteShader and glDeleteProgram
when deleting an object which does not exist,
INVALID_VALUE error should be generated
Change-Id: I52de29f7f980d63f0d35c9dfbfd3da17f89436e6
There was a bug in uncompress of GL_PALETTE4_RGB5_A1_OES and
GL_PALETTE8_RGB5_A1_OES texture formats.
They were considered RGB (not RGBA) for some reason.
This also fixed the pink sky in Wave Blazer Lite
Change-Id: I8c047d624d9bf09e2a0de7f7340d676cfe5e76c9
Copied the etc1 implementation from frameworks/base/opengl/libs/ETC1
to the translator common library, this is for keeping the translator
be independant.
Added support for ETC1 compressed textures in both GLESv1 and GLESv2
which is needed for some applications. (Cordy).
Change-Id: Ie615ed82896cb36c42fa640b83d9e93b723b3b80
