412901 - getComputedStyle should return actual (rounded) values for border widths

Status: RESOLVED FIXED

(Core :: DOM: CSS Object Model, defect, P1)

Description

[Robert O'Callahan (:roc) (email my personal email if necessary)]

We agreed in email this was a reasonable and useful thing to do.

Comment 1

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Attachment: fix

Comment 2

[Robert O'Callahan (:roc) (email my personal email if necessary)]

For the record, the reason this matters is jresig's testcase here:
http://dev.jquery.com/~john/ticket/ff3-offset/test/

He writes:

---------------------

There has been a change in Firefox 3 concerning the computation of borders and parent elements. For example, in Firefox 2 the result would've been 56 - in Firefox 3 it's now 57. This is fine - 57 is what all other browsers return.

However, both Firefox 2 and 3 return precise values for computed style values, meaning that if you set a border to a em or % value it will return a value up to multiple decimal places. No other browsers do this (they all round off to the nearest whole number).

The crux of the problem is that when trying to compute the width of an element (offsetWidth - computedLeftBorder - computedRightBorder) you get different values inbetween Firefox 3 and all other browsers, for example:

IE6, Opera9, Safari2, Safari3: 30 == 30: 56 - 13 - 13
FF2: 30.333399999999997 == 30 (rounds down to 30): 57 - 13.3333 - 13.3333
FF3: 29.333399999999997 == 30 (rounds down to 29): 56 - 13.3333 - 13.3333

---------------------

Basically, offsetWidth return the (rounded) actual border-box width, so if you want to determine the content width from it, you need to substract the actual border widths, and apparently Web authors expect getComputedStyle to return those widths. When it doesn't, you get the wrong result, as he describes.

[Note that you don't want to use offsetWidth anymore, you want to use getBoundingClientRect().width (when available in browsers) which returns more accurate (fractional) results in Gecko 1.9/FF3.]

Comment 3

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

These aren't used values; they're actual values.  What we're currently returning are used values.

Comment 4

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Oh, so we use the actual values, not the used values? OK :-)

Comment 5

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Comment on attachment 297708 [details] [diff] [review]
fix

r+sr=dbaron.

Could you also add a test that the border-width computes (getComputedStyle) to 0 if border-style is 'none' or 'hidden'?  (Maybe a second box in your existing test?)

Comment 6

[Robert O'Callahan (:roc) (email my personal email if necessary)]

checked in with updated test