[Webkit-unassigned] [Bug 156684] rem in media queries should be calculated using font-size:initial, not root element font-size

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=156684

--- Comment #9 from Myles C. Maxfield <mmaxfield at apple.com> ---
Hang on, computeNonCalcLengthDouble() is used for all style resolution, not just media queries. This is a situation where, if a media-query says "60rem" it refers to a different value than if an element says "60rem".

So we can't just modify computeNonCalcLengthDouble(). We have to set some state inside CSSToLengthConversionData which represents whether or not we are trying to calculate a media query or a regular style on an element.

Style::fontSizeForKeyword() is sensitive to the user's preferences, namely settings().defaultFixedFontSize() and settings().defaultFontSize(). Media queries probably should be sensitive to these preferences as well, which means we should use the logic inside Style::fontSizeForKeyword().

You're right about not having access to a document inside CSSPrimitiveValue, and this is desirable - individual CSS values shouldn't know what a whole Document is. Luckily, the CSSToLengthConversionData is created in MediaQueryEvaluator::evaluate() which has access to m_document. So we should interrogate m_document inside MediaQueryEvaluator::evaluate() and use our findings to set some new fields (which we will create) inside CSSToLengthConversionData. Then, computeNonCalcLengthDouble() can read those settings.

We should make sure that when we test this, we test all the relevant combinations of defaultFontSize(), defaultFixedFontSize(), media queries, and style on regular elements.

Does that help you understand this?

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-unassigned/attachments/20180301/68d2057f/attachment.html>