implement new negative flexing algorithm

[Tony Chang]

Correct formula for the desired result:

	space-shortage = sum(flex-basis) - available-space
	shrink-factor = space-shortage / sum(flex-basis * negative-flex))
	main-size = flex-basis * (1 - shrink-factor * negative-flex))


See http://lists.w3.org/Archives/Public/www-style/2012May/0291.html for a test case and some rationale.

[Tony Chang]

Created attachment 144151 [details]
Patch

[Tony Chang]

The spec reflects this change as well.

[Tony Chang]

Comment on attachment 144151 [details]
Patch

Actually, I can overflow and get into an infinite loop.  Let me fix that first.

[Tony Chang]

Created attachment 144161 [details]
Patch

[Ojan Vafai]

Comment on attachment 144161 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=144161&action=review

> Source/WebCore/rendering/RenderFlexibleBox.cpp:876
> +            if (availableFreeSpace > 0 && totalPositiveFlexibility > 0 && flexSign == PositiveFlexibility && isfinite(totalPositiveFlexibility))
>                  childSize += lroundf(availableFreeSpace * child->style()->positiveFlex() / totalPositiveFlexibility);
> -            else if (availableFreeSpace < 0 && totalNegativeFlexibility > 0  && flexSign == NegativeFlexibility)
> -                childSize += lroundf(availableFreeSpace * child->style()->negativeFlex() / totalNegativeFlexibility);
> +            else if (availableFreeSpace < 0 && totalWeightedNegativeFlexibility > 0 && flexSign == NegativeFlexibility && isfinite(totalWeightedNegativeFlexibility))

Might want to mention this isfinite check in the ChangeLog description.

[Tony Chang]

Created attachment 144164 [details]
Patch for landing

[WebKit Review Bot]

Comment on attachment 144164 [details]
Patch for landing

Clearing flags on attachment: 144164

Committed r118600: <http://trac.webkit.org/changeset/118600>

[WebKit Review Bot]

All reviewed patches have been landed.  Closing bug.