First, let me say, I am not sure of the proper place to post this issue. I tried recently to load a very large json file that details the world in which I wanted to view. The file is about 148 Megabyte. I have 16 GB of RAM and use on average about 16-25% of the memory tops. Therefore, I know this is not a memory issue on my computer. I have loaded the file locally so I know that it is not a network issue either. However, when I try to load and display the world, Chrome just crashes every single time. I was using Three.js as the platform for loading, so it may be a Three.js issue also / instead. I searched for help and it seems that various people are having difficulties loading large json files and displaying them as a world, despite the WebGL platform they are using. Any help or known fixes would be appreciated.

And where should the TM sign go? Do we need to ask Khronos to register a new trademark?

The WebGL repository is now being pulled wholesale into the Chromium repository for testing purposes. Chromium requires that all third-party files pass the Linux licensecheck utility. Currently many files in the WebGL repository fail this check, either because of missing or unrecognized per-file licenses.

A separate bug regarding the html5lib library has been filed at html5lib/html5lib-python#125 . The complete list of files failing licensecheck follows. Ideally all of these files will be fixed.

'third_party/webgl/src/sdk/demos/google/san-angeles/gles/bob/compare.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/sdk/demos/google/san-angeles/gles/t/build.sh' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/sdk/demos/google/san-angeles/gles/t/include/GLES/egltypes.h' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/sdk/demos/google/san-angeles/gles/t/include/GLES/egl.h' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/sdk/demos/google/san-angeles/gles/t/include/GLES/gl.h' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/sdk/demos/google/san-angeles/gles/t/app-linux.c' has non-whitelisted license 'MIT/X11 (BSD like) LGPL (v2.1 or later)'
'third_party/webgl/src/sdk/demos/google/san-angeles/gles/build.sh' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/sdk/demos/google/san-angeles/gles/include/GLES/egltypes.h' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/sdk/demos/google/san-angeles/gles/include/GLES/egl.h' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/sdk/demos/google/san-angeles/gles/include/GLES/gl.h' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/sdk/demos/google/san-angeles/gles/app-linux.c' has non-whitelisted license 'MIT/X11 (BSD like) LGPL (v2.1 or later)'
'third_party/webgl/src/sdk/demos/google/san-angeles/glut/app-glut.c' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/sdk/demos/google/san-angeles/glut/build.sh' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/sdk/demos/google/san-angeles/glut/include/GLES/egltypes.h' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/sdk/demos/google/san-angeles/glut/include/GLES/egl.h' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/sdk/demos/google/san-angeles/glut/include/GLES/gl.h' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/sdk/demos/google/san-angeles/glut/app-linux.c' has non-whitelisted license 'MIT/X11 (BSD like) LGPL (v2.1 or later)'
'third_party/webgl/src/sdk/demos/google/nvidia-vertex-buffer-object/VertexBufferObject2.java' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/specs/1.0/extract-idl.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/specs/latest/1.0/extract-idl.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/specs/latest/2.0/extract-idl.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/specs/1.0.2/extract-idl.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/index.php' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/inputstream.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/html5parser.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/treewalkers/pulldom.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/treewalkers/etree.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/treewalkers/lxmletree.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/treewalkers/init.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/treewalkers/soup.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/treewalkers/dom.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/treewalkers/_base.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/treewalkers/simpletree.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/treewalkers/genshistream.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/sanitizer.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/ihatexml.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/filters/inject_meta_charset.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/filters/lint.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/filters/sanitizer.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/filters/optionaltags.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/filters/_base.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/filters/whitespace.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/filters/formfiller.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/treebuilders/etree.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/treebuilders/etree_lxml.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/treebuilders/init.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/treebuilders/soup.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/treebuilders/dom.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/treebuilders/_base.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/treebuilders/simpletree.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/utils.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/tests/test_stream.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/tests/runparsertests.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/tests/test_treewalkers.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/tests/tokenizertotree.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/tests/test_sanitizer.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/tests/test_serializer.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/tests/test_whitespace_filter.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/tests/init.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/tests/test_encoding.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/tests/test_formfiller.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/tests/support.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/tests/test_tokenizer.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/tests/test_parser.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/tests/performance/concatenation.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/tests/runtests.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/tests/mockParser.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/serializer/htmlserializer.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/serializer/xhtmlserializer.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/serializer/init.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/init.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/constants.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/resources/html5lib/src/html5lib/tokenizer.py' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/extensions/atom.xsl' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/extensions/schematron/iso_dsdl_include.xsl' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/extensions/schematron/webgl_message_patch.xsl' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/extensions/schematron/iso_svrl_for_xslt1.xsl' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/extensions/standards.xsl' has non-whitelisted license 'UNKNOWN'
'third_party/webgl/src/extensions/rev_utils.xsl' has non-whitelisted license 'UNKNOWN'

Stephen Adams from Google has pointed out that the tests for typed arrays' set() methods are insufficient. In particular, the tests must verify the case where set() is called with a different type of typed array view than the receiver's, due to how Web IDL's overloads work. (The original intent was that one could only pass a Float32Array to Float32Array.set, but it was discovered some time ago that a different view type like Uint8Array will match the overload taking "type[]".)

Fill in the section listing differences between WebGL 2.0 and OpenGL ES 3.0

Hi,

Not sure that I'm on right place... but still.
When using the extension oes-element-index-uint and when the index buffer is set with DYNAMIC_DRAW, sometimes the oes-element-index-uint test doesn't pass.

I discover this problem when I was developing my app:

• here's with STATIC index buffer
• here's with DYNAMIC index buffer (sometimes the sphere just doesn't appear).

When the test doesn't pass, refreshing the page fix it everytime...
What are the conditions to trigger the fail? hmm.. good question, I usually open a new tab and load the same page and eventually one will fail (sometimes a lot of consecutive tab works, and sometimes they fail..).
But it usually often fails when restarting chrome and loading the page.

drawElements trigger this error :

WebGLRenderingContext: GL ERROR :GL_OUT_OF_MEMORY : glFramebufferTexture2D: <- error from previous GL command 
WebGLRenderingContext: GL ERROR :GL_OUT_OF_MEMORY : glTexImage2D: <- error from previous GL command 

ps: It always works (so far for me) when using STATIC for the index buffer
ps2 : firefox aurora support oes-element-index-uint and the problem also appears
ps3: Tested on Windows 7, my graphic card: ATI Mobility Radeon HD 5650, I didn't test it yet on another computer... so maybe it's specific to my computer...

GLSL ES spec 1.0.17 section 10.13 says that not writing gl_Position results in undefined behavior, which is undesirable in WebGL. Current WebGL implementations do seem to enforce statically writing to gl_Position, but it's not mentioned in the WebGL spec.

There's also a related problem in the conformance test glsl/literals/float_literal.vert.html, where vertex shader does not write gl_Position. I think writing to gl_Position should simply be added there.

A new test could also be written that specifically checks that not writing to gl_Position in a vertex shader will result in a shader compilation error.

I recently upgraded to OS X Mavericks, and "make" fails in the extensions directory.

The error is:
./validate EXT_color_buffer_half_float
xsltproc --stringparam basepath echo EXT_color_buffer_half_float/ | sed -e "s/[^\/]*/../g"
extension.xsl EXT_color_buffer_half_float/extension.xml > EXT_color_buffer_half_float/index.html
compilation error: file extension.xsl line 56 element attribute
xsl:attribute: The attribute name 'xmlns' is not allowed.
make: *** [EXT_color_buffer_half_float/index.html] Error 5

I believe the problem is that we're using an HTML doctype (), but attempting to to set an xml namespace for XHTML.

We should probably pick one? HTML with no xml namespace, or use an XHTML namespace.

Here's my version of xsltproc (output from "xsltproc --version"):
Using libxml 20900, libxslt 10128 and libexslt 817
xsltproc was compiled against libxml 20900, libxslt 10128 and libexslt 817
libxslt 10128 was compiled against libxml 20900
libexslt 817 was compiled against libxml 20900

We need text for the WebGLSync interface

It was reported on the public mailing list that an
implementation was failing with GLSL code such as this:

uniform float x, y;

We should add a test that explicitly exercises this.

The GL spec says "glBindAttribLocation can be called before any vertex shader objects are bound to the specified program object. It is also permissible to bind a generic attribute index to an attribute variable name that is never used in a vertex shader."

This should be exercised in the test suite, since all (?) WebGL implementations implement a variable name translation via ANGLE when compiling shaders. This issue was discovered when running some Unity examples in WebKit.

The test would have to call bindAttribLocation on a number of attributes before the compilation step.

We need text for the WebGLVertexArrayObject interface

Chris Green from Imagination has pointed out that the following tests introduced in the 1.0.2 conformance suite may have bugs:

conformance/ogles/GL/biuDepthRange/biuDepthRange_001_to_002.html
conformance/ogles/GL/gl_FragCoord/gl_FragCoord_001_to_003.html

See https://www.khronos.org/bugzilla/show_bug.cgi?id=840 .

Propose API for mapBufferRange equivalent -> getBufferSubData

Note there are a bunch of enums exposed by mapBufferRange that will need updating too (if only to remove the comment)

The conformance test typedarrays/data-view-test.html
explicitly checks to see if
var v = DataView(new ArrayBuffer);
throws an exception.

This used to be the correct behaviour, but it has since changed. The ECMAScript 6 Draft Specification allows DataView to be used as a function returning a constructed DataView object.

http://wiki.ecmascript.org/doku.php?id=harmony:specification_drafts

I'll submit a pull request to change the test in ToT, but we still need to do something about the 1.0.2 test suite. WebKit currently fails the test even though it is doing the right thing. But Chromium is passing the test, so updating it will cause breakage there.

My suggestion is to remove just this part of the test for 1.0.2, but leave in the updated part for ToT.

We have a dodgy redirect to 1.0. Instead we should have a nice landing page that explains 1.0 vs 2.0.

Just changed the dead SVN links to the github repo. 669f2d1

The gl-disabled-vertex-attrib.html conformance test incorrectly assumes that vertex output varyings remain exactly the same when passed to the fragment shader. This is not guaranteed, due to perspective correction.

Most implementations still pass the test because their implementation of the reciprocal operation used for perspective correction is exact (infinite precision) for 1.0. Some implementations don't make that guarantee, and the spec appears to allow for such finite precision for any input. This test could fail on any implementation by moving the quad that is being rendered back or forth.

This test should not depend on the perspective correction implementation. I believe it can be fixed by moving the isCorrectColor test from the fragment shader into the vertex shader.

ESSL 1.00.17 states:
"Shared globals are variables that can be accessed by multiple compilation units. In GLSL ES the only
shared globals are uniforms. Varyings are not considered to be shared globals since they must pass
through the rasterization stage before they can be read by the fragment shader."
"Shared globals must have the same name, storage and precision qualifiers."

The TeapotPerPixel demo breaks this rule since it shares uniforms between the vertex and fragment shader, but uses the default precision (highp) in the vertex shader and mediump in the fragment shader.

(It possibly works on many (desktop) implementations since the precision qualifiers may just be ignored there.)

This is a copy of the bugzilla issue:

http://www.khronos.org/bugzilla/show_bug.cgi?id=729

"The "oes-texture-float-and-half-float-linear" is failing because it is calling glGenerateMipmap() when a float texture is bound. That call tries to render to a float buffer, which will fail in our case. These tests should be handle the case where this mipmap generation fails."

I recently ran into an issue where in the vertex shader I had this

// -- vertex shader --
...
attribute vec4 color;
varying vec4 v_color
void main() {
  v_color = color;
  ...
}

// -- fragment shader --
...
uniform vec4 v_color
void main() {
  gl_FragColor = v_color;
}

I got no error. The error is in the vertex shader I used 'varying" but in the fragment shader I used "uniform". It's not clear to me if this should generate an error or not.

In #340 the WebGL conformance suite harness for running the imported ES 3.0 tests was relaxed to be identically as strict as ES 3.0's GTF harness, because of problems with unreferenced uniforms being optimized away by some GLSL compilers. This was an expedient workaround for allowing the WebGL 1.0.2 conformance suite to be passed.

A better fix would be to fix the tests and make the ES 3.0 conformance suite harness stricter, requiring that all referenced uniforms be present. At that point the WebGL harness should me made stricter as well.

This is a follow-on to Issue 236.

sdk/demos/google/resources/OESVertexArrayObject.js is an attempt to emulate the OES_vertex_array_object extension 100% in JavaScript.

One thing that's missing is that it must not let buffers actually get deleted unless they are no longer referenced by any vertex array objects.

We need text for the WebGLSampler interface

Some of the tests under https://www.khronos.org/registry/webgl/sdk/tests/extra fail with syntax errors:

"Uncaught SyntaxError: In strict mode code, functions can only be declared at toplevel or immediately within another function."

• https://www.khronos.org/registry/webgl/sdk/tests/extra/fbo-lost-context.html
• https://www.khronos.org/registry/webgl/sdk/tests/extra/out-of-memory.html
• https://www.khronos.org/registry/webgl/sdk/tests/extra/out-of-resources.html
• https://www.khronos.org/registry/webgl/sdk/tests/extra/out-of-vram.html

"Uncaught ReferenceError: canvas is not defined"

• https://www.khronos.org/registry/webgl/sdk/tests/extra/offscreen-issue.html

Filed this on the Khronos tracker: https://www.khronos.org/bugzilla/show_bug.cgi?id=890 but I noticed that this one seems more active so filing here as well.

The WebGL 1.0.2 test glsl/misc/shader-with-for-loop.html has "precision highp float" in the shader. highp floating point support is optional according to the spec, so this line should be removed.

This causes the test to fail on some mobile devices, see for example Chrome on Tegra 3.

If you run https://www.khronos.org/registry/webgl/sdk/tests/conformance/canvas/buffer-offscreen-test.html directly it passes in every browser.

If you run the same test by going to https://www.khronos.org/registry/webgl/sdk/tests/webgl-conformance-tests.html and press the "run" button for that test, it fails 1 out of 9 tests on firefox for OSX or Linux (strangely it passes both ways on windows).

on slow machines conformance/uniforms/uniform-samplers-test.html can bring up "page is unresponsive" types of message.

Solution: yield once in awhile.

It would be nice to refactor https://www.khronos.org/registry/webgl/sdk/debug/webgl-debug.js to use the WEBGL_lose_context extension where it exists.

Specifically webgl-debug.js contains code to emulate context lost events and help debug them by throwing and specific times. That code though does not handle extensions in general. Using WEBGL_lose_context would solve that for implementations that support WEBGL_lose_context.

running test-runner with --version 1.0.0 --browser FirefoxWinOpenGL will result in an immediate crash in firefox 28

Subsequent runs would show "This browser does not appear to support WebGL"

They spec says every one past N is GL_NONE. Make sure the test for drawBuffers checks this.

We need to fill in the one line explanation for all the API entry points, and link back to the ES spec.

Currently the test-runner is designed mainly for GPU vendors to test their drivers. However, it would be great to enhance it to generate WebGL conformance suite submissions for browser vendors.

The following changes would need to be made in order to do this:

• Add an optional command line argument to re-enable the GPU driver bug workarounds the browser may have. (The default must continue to be to disable these workarounds.)
• The harness would need to gather OS, GPU and graphics driver version information and add them to the results.

The http://www.khronos.org/webgl/ site will need some new links, as well as an update to the explanatory text. That content is not in this repository.

Looking at sdk/tests/conformance/renderbuffers/framebuffer-object-attachment.html , it seems tests are needed for the following situations:

1. allocating a DEPTH_STENCIL renderbuffer with renderbufferStorage, and then attaching it to a framebuffer
2. attaching a renderbuffer to a framebuffer's DEPTH_STENCIL attachment, and then allocating it with renderbufferStorage

On platforms not supporting OES_packed_depth_stencil, an implementation emulating the DEPTH_STENCIL attachment point with two separate renderbuffers may well get case (2) wrong. A rendering test is needed ensuring that the depth and stencil buffers are working properly for these two cases.

http://www.khronos.org/registry/webgl/specs/latest/2.0/#3.7 says:

[NoInterfaceObject]
interface WebGL2RenderingContextBase implements WebGLRenderingContextBase
{
...
}

WebIDL does not have this notion of implements. It is not clear what the intent is but you probably meant something along the lines of

interface WebGL2RenderingContextBase {
  ...
}
WebGL2RenderingContextBase implements WebGLRenderingContextBase;

which means that all the members from WebGLRenderingContextBase are also members of WebGL2RenderingContextBase.

The majority of tests assume that the attribute locations are in the same order as defined by the shader source. I recently enabled symbol mangling in WebKit, and have noticed that (rarely) the indexes are not in the same order.

We should probably update the tests, since order is not guaranteed, although I expect to see some broken content in WebKit due to this. I'm not sure how to best address this.

In his work on supporting lost contexts for 2D Canvas, Justin Novosad pointed out that the WebGL spec isn't clear that WebGLContextEvent initializes the cancelable flag to true upon creation. Event.preventDefault() just sets the canceled flag, and that can only be done if the event is cancelable, so this is worth clarifying.

He also pointed out that the 'webglcontextrestored' event is unnecessarily cancelable in Chrome's implementation, but I'm not sure this matters one way or the other since there's no default behavior, unlike the 'webglcontextlost' event. Both should probably be cancelable, just by virtue of WebGLContextEvent being cancelable.

From the tip conformance test page:
www.khronos.org/registry/webgl/sdk/tests/webgl-conformance-tests.html

There's a version drop-down, but it doesn't seem to change which version of the tests are run.

glValidEnumContexts in https://www.khronos.org/registry/webgl/sdk/debug/webgl-debug.js is out of date.

It was copied from a function long before WebGL 1.0 shipped. It's also missing compressedTexImage2D and compressedTexSubImage2D

I would like to join in this group, may you add me?

I need to write more scissor and viewport tests.

Specifically,

• that neither scissor nor viewport change when anything about the current
  framebuffer changes in including adding/removeing attachments, deleting
  attachments, changing the size of attachments, changing which fbo
  is bound. Some of these tests exist.
• that scissor can be anything that is not a negative width or negative height.
  In other words

scissor(-100000, -20000, 300000, 400000)

is valid

• That viewport can be set larger than the current backbuffer / fbo and that larger
  setting correctly effects rendering.

We need text for the WebGLTransformFeedback interface

WebGL is removing verticies after afew seconds of rendering.

On my computer, It starts out fine, like the first image below, but within a few seconds, becomes the mess in the second part of that image.

you can see that it works in your browser here

I've tried clearing my history, cache, cookies, anything I can think of. I've even switched Firefox from Chrome. nothing seems to work.

I have an ATI Radeon™ HD 3400 Series graphics card, and I'm on Windows 7 using up to date chrome. Any idea of what the issue could be?

Recently a bug was reported by the three.js and Verold.com developers affecting Qualcomm's Adreno GPUs. Computation of fog using an expression involving gl_FragCoord.z and gl_FragCoord.w causes incorrect compilation of the fragment shader and incorrect rendering.

A WebGL conformance test should be added for this, despite the fact that this will prevent the conformance suite from passing on Qualcomm GPUs. It can be added after the 1.0.3 conformance suite is snapshotted.

This probably needs some spec updates to make it clear what happens. The simple version is all extensions except WEBGL_lose_context need to be explicitly re-enabled after a context is lost/restored.

So for example

ext1 = gl.getExtension("OES_vertex_array_object");
... // context lost, context restored
ext1.bindVertexArrayOES();  // is supposed to be a no-op 
                            // even though the context is restored.

According to the spec

ext1 = gl.getExtension("OES_vertex_array_object");

Should return the same object which probably means implementation-wise each extension needs to track context lost itself

We do not have coverage for declaring empty arrays. We should test both:

float f[];
float f[0];

We already have tests for negative size inside ogles/GL/build/array*

Fill in the section that lists other differences between WebGL2 and WebGL1

We need text for the WebGLQuery interface.