Bug 199789 – [WebGL] Remove software rendering and simplify context creation on macOS

• Component: New Bugs
• Status: RESOLVED FIXED

[Dean Jackson]

[WebGL] Remove software rendering and simplify context creation on macOS

[Dean Jackson]

Created attachment 374096 [details]
Patch

[Sam Weinig]

Comment on attachment 374096 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=374096&action=review

> Source/WebCore/ChangeLog:9
> +        We don't ever want to fall-back to the software renderer. We'd be better
> +        off failing to create the context completely.

Why is this? Why did we support it in the first place?

> Source/WebCore/platform/graphics/GraphicsContext3DManager.cpp:177
> -    ASSERT(m_contexts.contains(context));
> +    if (!m_contexts.contains(context))
> +        return;

What is this change for?

> Source/WebCore/platform/graphics/GraphicsContext3DManager.cpp:212
> +    if (!context)
> +        return;

What is this change for?

[Dean Jackson]

Comment on attachment 374096 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=374096&action=review

>> Source/WebCore/ChangeLog:9
>> +        off failing to create the context completely.
> 
> Why is this? Why did we support it in the first place?

Because we thought it was a good idea, but the performance and conformance are not acceptable.

>> Source/WebCore/platform/graphics/GraphicsContext3DManager.cpp:177
>> +        return;
> 
> What is this change for?

If we have a null m_contextObj, we didn't add ourselves to the GC3DManager. However, the GC3D was still created, so it tries to remove itself from the manager at destruction time.

I should have noted this in the ChangeLog - I'll add it.

>> Source/WebCore/platform/graphics/GraphicsContext3DManager.cpp:212
>> +        return;
> 
> What is this change for?

This shouldn't be needed.

[Dean Jackson]

Committed r247453: <https://trac.webkit.org/changeset/247453>

[Radar WebKit Bug Importer]

<rdar://problem/53125394>