This avoids grabbing the document when values that inherit from the body,
whatever that means, aren't under the body.
In that case we'll get a semi-random value, but that's also mishandled by Gecko
anyways (and probably Blink, though haven't tested), and doesn't really make
much sense.
First, we define computed::CSSPixelLength which contains a CSSFloat, a
pixel value, and then we replace computed::Length with CSSPixelLength.
Therefore, the |ComputedValue| of NoCalcLength, AbsoluteLength,
FontRelativeLength, ViewportPercentageLength, CharacterWidth, and
PhysicalLength is CSSPixelLength.
Besides, we drop NonNegativeAu, and replace computed::NonNegativeLength
with NonNegative<computed::Length>. (i.e. NonNegative<CSSPixelLength>)
Reduce usage of fmt in serialization and error reporting
`format!` and `write!` create a somewhat-heavyweight `Formatting` struct and use dynamic dispatch to call into impls of `Dispaly` and related traits. The former also allocates an intermediate string that is sometimes unnecessary.
I started looking into this from https://bugzilla.mozilla.org/show_bug.cgi?id=1355599, but I expect the impact there will be small to insignificant. It might be a slightly less so on parsing (error reporting).
<!-- Reviewable:start -->
---
This change is [<img src="https://reviewable.io/review_button.svg" height="34" align="absmiddle" alt="Reviewable"/>](https://reviewable.io/reviews/servo/servo/18355)
<!-- Reviewable:end -->
Gecko currently doesn't support calc inside media queries. We should
also remove the calc support temporarily for parity with gecko. We can
add this support back in next releases.
Use whether we've computed any viewport unit instead.
This is more accurate (we avoid restyling unnecessarily if we've found anything
ever on the stylesheet, but that hasn't matched).
This has the benefit of also matching Gecko, and simplify some code and
fishyness around, and also hopefully speeding up stylesheet parsing.
Right now when calling getComputedStyle with an element from another document,
we return the style using the pres context of that document, not of the document
of the window getComputedStyle was called on, so this holds.
But it will stop holding if we ever change this, so assert it doesn't happen.
Bug: 1374062
Reviewed-By: Manishearth
MozReview-Commit-ID: 3g8yQWWdsen
Zoom changes can change the meaning of ems, so we can't re-evaluate media
queries with the old values, because otherwise we may miss a restyle.
Bug: https://bugzilla.mozilla.org/show_bug.cgi?id=1357461
MozReview-Commit-ID: HQInvR7RPqR
If there are multiple prefixed/non-prefixed @keyframes with the same name;
* non-prefixed rule overrides earlier rules.
* prefixed rule overrides earlier prefixed rules.
To make it possible to check the rule type when parsing lengths, we need to pass
the `ParserContext` down through many layers to the place where length units are
parsed.
This change leaves it unused, so it's only to prepare for the next change.
MozReview-Commit-ID: 70YwtcCxnWw
