Bug 55022 – REGRESSION: Accelerated transitions are jumpy

• Component: Layout and Rendering
• Status: RESOLVED FIXED

[Simon Fraser (smfr)]

Go to http://www.eleqtriq.com/2010/05/css-3d-matrix-transformations/, and scroll down to the "Example: Coverflow in CSS" section, and click Open Example. In the example, alternately hover over two adjacent cards. Note how they jump backwards and forwards, rather than animating smoothly. This is a regression from Safari 5.

[Simon Fraser (smfr)]

<rdar://problem/9040742>

[Simon Fraser (smfr)]

We're getting the timing function wrong.

[Simon Fraser (smfr)]

Created attachment 83611 [details]
Patch

[Eric Seidel (no email)]

Comment on attachment 83611 [details]
Patch

rs=me.

[Simon Fraser (smfr)]

http://trac.webkit.org/changeset/79568

Comment on attachment 83611 [details]
Patch

Consider getting rid of the text in the test in the success case and landing the results in a cross-platform location.

[Simon Fraser (smfr)]

I plan to do a pass over the transition/animation tests and make them dump as text.

However, the pixel results aren't cross platform; the CALayer can paint non pixel-aligned.

[Dean Jackson]

sorry! my mistake!

[Dean Jackson]

I take it back. I'm not sorry. This was Chris :)

[Chris Marrin]

It's still wrong on the Win version. I'll add a patch for that presently.

[Chris Marrin]

Created attachment 84257 [details]
Patch botched by Windows

[Chris Marrin]

Created attachment 84258 [details]
Windows Patch

[Simon Fraser (smfr)]

Comment on attachment 84258 [details]
Windows Patch

I think we should do the null check higher up and always pass in a TimingFunction here, because now these magic numbers are in 3 places.

[Chris Marrin]

Created attachment 84482 [details]
Patch

[Simon Fraser (smfr)]

Comment on attachment 84482 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=84482&action=review

> Source/WebCore/platform/graphics/ca/win/PlatformCAAnimationWin.cpp:128
> +    ASSERT(timingFunction);
> +    

The Mac version has no blank line after the assert.

[Chris Marrin]

Committed r80266: <http://trac.webkit.org/changeset/80266>

[Adam Roben (:aroben)]

Comment on attachment 84482 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=84482&action=review

> Source/WebCore/platform/animation/TimingFunction.h:109
> +    static const CubicBezierTimingFunction* defaultTimingFunction()
> +    {
> +        DEFINE_STATIC_LOCAL(const CubicBezierTimingFunction, dtf, ());
> +        return &dtf;
> +    }

This is wacky, and could lead to assertion failures due to adoptRef never being called. I'd suggest writing it like this:

static const CubicBezierTimingFunction* function = create().leakRef();
return function;

[Chris Marrin]

Created attachment 84623 [details]
Patch changing TimingFunction per Adam's suggestion

[Chris Marrin]

Committed r80289: <http://trac.webkit.org/changeset/80289>