[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Public WebGL] Non-passable 1.0.1 conformance test: relying on canvas2D drawImage to not do colorspace conversion



----- Original Message -----
> Hi Benoit,
> 
> Thanks for finding this bug and for the fix. You're right, these two
> tests were making incorrect assumptions.
> 
> The 1.0.1 conformance suite won't be considered "done" until we have
> multiple browsers passing it on multiple platforms. For this reason
> I've gone ahead and merged your bug fix back to the 1.0.1 snapshot.

Great, I didn't realize we could still make fixes until it's known to be passing on multiple implementations. That's a great rule.

Sorry about the const keyword, I didn't realize that it was nonstandard. I've moved the 'red' and 'green' variables back to global scope though, fixes the timeouts.

Benoit

> 
> -Ken
> 
> 
> On Wed, Apr 18, 2012 at 2:20 PM, Benoit Jacob <bjacob@mozilla.com>
> wrote:
> >
> > Hi,
> >
> > I've just checked in a fix on the development branch (r17499):
> >
> > https://cvs.khronos.org/svn/repos/registry/trunk/public/webgl/sdk/tests/conformance/textures/tex-image-and-sub-image-2d-with-image.html
> >
> > This test was non-passable (specifically, runTest2 and runTest3)
> > because it was taking a color-managed PNG image, drew it onto a 2D
> > canvas, obtained the data URL of this canvas, used it as the
> > source for a WebGL texture, and was expecting exact pixel values
> > from this texture. The problem was that even though
> >
> >    gl.pixelStorei(gl.UNPACK_COLORSPACE_CONVERSION_WEBGL, gl.NONE);
> >
> > was used, it doesn't help with the fact that as soon as the image
> > was drawn onto the 2D canvas, it was colorspace-transformed and
> > the canvas doesn't contain the full information of the original
> > image before colorspace transformation; so there is no hope that
> > the original non-colorspace-transformed image could be obtained
> > exactly from the 2D canvas.
> >
> > The question is what can we do about 1.0.1?
> >  - I assume that we don't want any tweaks anymore in 1.0.1?
> >  - could be make a 1.0.1.1? Or call that 1.0.2 and say that the
> >  development version is now heading towards 1.1?
> >
> > Notice that there is another way to fix this bug, without changing
> > any JS code: we could use a non-color-managed version of
> > red-green.png, as found here (courtesy of the libpng maintainer):
> >
> > http://people.mozilla.org/~bjacob/red-green-stripped.png
> >
> > This makes the test pass on colorspace-corrected systems without
> > code changes.
> >
> > Cheers,
> > Benoit
> >
> > -----------------------------------------------------------
> > You are currently subscribed to public_webgl@khronos.org.
> > To unsubscribe, send an email to majordomo@khronos.org with
> > the following command in the body of your email:
> > unsubscribe public_webgl
> > -----------------------------------------------------------
> >
> 

-----------------------------------------------------------
You are currently subscribed to public_webgl@khronos.org.
To unsubscribe, send an email to majordomo@khronos.org with
the following command in the body of your email:
unsubscribe public_webgl
-----------------------------------------------------------