Bug 94300 – [GTK] three.js example webgl_terrain_dynamic.html doesn't work

• Component: WebGL
• Status: RESOLVED DUPLICATE

[Tomeu Vizoso]

The animation doesn't appear once the textures finish loading, here is a link for easier testing:

http://mrdoob.github.com/three.js/examples/webgl_terrain_dynamic.html

It works fine on Chrome.

[Tomeu Vizoso]

Created attachment 162460 [details]
Patch

[Martin Robinson]

Comment on attachment 162460 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=162460&action=review

> Source/WebCore/ChangeLog:11
> +        No new tests (OOPS!).

This change really needs a test of some kind, I think.

[Tomeu Vizoso]

(In reply to comment #2)
> (From update of attachment 162460 [details])
> View in context: https://bugs.webkit.org/attachment.cgi?id=162460&action=review
> 
> > Source/WebCore/ChangeLog:11
> > +        No new tests (OOPS!).
> 
> This change really needs a test of some kind, I think.

It's probably take a while until I find time to do that, in case someone else wants to take this and push it forward.

[Martin Robinson]

I can confirm that this fixes http://hexgl.bkcore.com as well.

[arno.]

(In reply to comment #3)
> (In reply to comment #2)
> > (From update of attachment 162460 [details] [details])
> > View in context: https://bugs.webkit.org/attachment.cgi?id=162460&action=review
> > 
> > > Source/WebCore/ChangeLog:11
> > > +        No new tests (OOPS!).
> > 
> > This change really needs a test of some kind, I think.
> 
> It's probably take a while until I find time to do that, in case someone else wants to take this and push it forward.

I've already isoled a minimal testcase (which I can't upload now since it's on another computer). I'll try to evolve it into a proper test new week.

[arno.]

Created attachment 163454 [details]
testcase

Hi,

here is a testcase.

when I put it in LayoutTests/fast/canvas/webgl, and display it with
Tools/Scripts/run-launcher --gtk file:///home/arno/webkit/WebKit.upstream/LayoutTests/fast/canvas/webgl/webgl-doubleframebuffer.html --enable-webgl

I can check that it fails on trunk but suceeds with your patch.

Unfortunately, when I try to run it with ./Tools/Scripts/run-webkit-tests --gtk --force fast/canvas/webgl/webgl-doubleframebuffer.html it fails, so can you check this test it on your systems ?



---
About the test failure, there is the following error

CONSOLE MESSAGE: line 24: *** Error loading shader '[object WebGLShader]':INVALID_OPERATION
CONSOLE MESSAGE: line 24: *** Error compiling shader '[object WebGLShader]':
CONSOLE MESSAGE: line 239: TypeError: 'null' is not an object (evaluating 'gl.createBuffer')
 FAIL successfullyParsed should be true (of type boolean). Was undefined (of type undefined).



actually, gl.getError() is already INVALID_OPERATION just after gl.getContext()
This is true also in other webgl tests, but it does not trigger an exception (because the first gl.getError() test appear at other moments). My guess is, it could be a bug of my driver (ati catalyst) with Xvfb.

[Martin Robinson]

(In reply to comment #6)

> actually, gl.getError() is already INVALID_OPERATION just after gl.getContext()
> This is true also in other webgl tests, but it does not trigger an exception (because the first gl.getError() test appear at other moments). My guess is, it could be a bug of my driver (ati catalyst) with Xvfb.

Perhaps the WebGL WebKitWebSetting is not enabled in DumpRenderTree?

[arno.]

(In reply to comment #7)
> (In reply to comment #6)
> 
> > actually, gl.getError() is already INVALID_OPERATION just after gl.getContext()
> > This is true also in other webgl tests, but it does not trigger an exception (because the first gl.getError() test appear at other moments). My guess is, it could be a bug of my driver (ati catalyst) with Xvfb.
> 
> Perhaps the WebGL WebKitWebSetting is not enabled in DumpRenderTree?

But the other tests run fine; and getContext returns a webgl context (with a non getError).
I could check with printf-debug that:
webgl settings are enabled during the test.
At some point (in WebGLRenderingContext::initVertexAttrib0), a buffer is created with:

    GLuint o = 0;
    glGenBuffers(1, &o);

And after glGenBuffers, o is 0 and glGetError is GL_NO_ERROR.
Following call to glBufferData will trigger an error because:
http://www.opengl.org/sdk/docs/man/xhtml/glBufferData.

GL_INVALID_OPERATION is generated if the reserved buffer object name 0 is bound to target.

[Martin Robinson]

*** This bug has been marked as a duplicate of bug 126538 ***