Bug 79169 – [Qt][WK2] Enable WebGL for WebKit2-Qt Linux/GLX

• Component: WebGL
• Status: RESOLVED DUPLICATE

[Noam Rosenthal]

see subject.

[Zeno Albisser]

noam: i'm reassigning this one to myself, i guess that is fine with you.
i already started to look into why webgl stopped working with QtTestBrowser. - As a starting point.

[Noam Rosenthal]

(In reply to comment #1)
> noam: i'm reassigning this one to myself, i guess that is fine with you.
> i already started to look into why webgl stopped working with QtTestBrowser. - As a starting point.

Great!

[Zeno Albisser]

Created attachment 152206 [details]
patch for review.

[Simon Hausmann]

Comment on attachment 152206 [details]
patch for review.

View in context: https://bugs.webkit.org/attachment.cgi?id=152206&action=review

> Source/WebCore/platform/graphics/qt/GraphicsContext3DQt.cpp:88
> +#if OS(DARWIN)
>      RefPtr<GraphicsSurface> m_frontBufferGraphicsSurface;
>      RefPtr<GraphicsSurface> m_backBufferGraphicsSurface;
> +#else
> +    RefPtr<GraphicsSurface> m_graphicsSurface;
> +#endif

I supposed that there are two users of GraphicsSurface: Those who need double-buffering and those who don't (they write to the surface only once and then are done with it).

I think should be encapsulated in GraphicsSurface itself (for example having a front and a back PlatformGraphicsSurface) and be a parameter of the creating API. Requiring OS(DARWIN) on the
_caller_ side feels wrong.

Is that the direction you plan to take after this patch or in another iteration?

> Source/WebCore/platform/graphics/texmap/TextureMapperBackingStore.cpp:71
> +    static_cast<TextureMapperGL*>(textureMapper)->drawTexture(m_textureID, 0, m_graphicsSurfaceSize, targetRect, adjustedTransform, opacity, mask);

Similarly here, perhaps drawTexture should be a function of GraphicsSurface that depending on the platform implementation uses drawTextureRectangleARGB or drawTexture of a TextureMapperGL parameter.

[Zeno Albisser]

(In reply to comment #4)
> I supposed that there are two users of GraphicsSurface: Those who need double-buffering and those who don't (they write to the surface only once and then are done with it).
> 
> I think should be encapsulated in GraphicsSurface itself (for example having a front and a back PlatformGraphicsSurface) and be a parameter of the creating API. Requiring OS(DARWIN) on the
> _caller_ side feels wrong.
> 
> Is that the direction you plan to take after this patch or in another iteration?
That's exactly what am working on right now.

[Noam Rosenthal]

Comment on attachment 152206 [details]
patch for review.

View in context: https://bugs.webkit.org/attachment.cgi?id=152206&action=review

A bit too #ifdeffy... i'd rather it would be capability based and not OS based.

>> Source/WebCore/platform/graphics/qt/GraphicsContext3DQt.cpp:88
>> +#endif
> 
> I supposed that there are two users of GraphicsSurface: Those who need double-buffering and those who don't (they write to the surface only once and then are done with it).
> 
> I think should be encapsulated in GraphicsSurface itself (for example having a front and a back PlatformGraphicsSurface) and be a parameter of the creating API. Requiring OS(DARWIN) on the
> _caller_ side feels wrong.
> 
> Is that the direction you plan to take after this patch or in another iteration?

I'd rather this be done with a runtime flag like GraphicsSurface::isDoubleBufferred, even now. Then we can keep m_front... and initialize the back one only if double buffered is not supported.

[Zeno Albisser]

*** This bug has been marked as a duplicate of bug 90881 ***