94855 - getComputedStyle; absolute and relative positioned elements

Status: RESOLVED FIXED

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

Description

[Stephan Tolksdorf]

I think the behaviour of Mozilla's implementation of DOM2's getComputedStyle is 
inconsistent.

Look at bug #93437 to read about the intended behaviour of getComputedStyle.

Presuming the following pseudo document:

<div id="layer1" style="position: absolute;">
  layer 1
  <div id="layer2" style="position: see below;">layer 2</div>
</div>

getComputedStyle will return the following values for layer2's "top"
property:
 - position set to "absolute":
        "top" relative to the containing block, ie. "top" relative to layer1

 - position set to "relative"
   or position set to "static":
         "top" relative to the viewport

If the position is set to "relative" shouldn't getComputedStyle return
zero?
If the position is set to "relative" shouldn't getComputedStyle return
"top" relative to where it would be if statically positioned?

Does the specified behaviour allow to findout the relative position of a 
statically positioned element with respect to its containing block
not using proprietary offsetLeft/Top properties)? [Missing feature?]

Comment 1

[Stephan Tolksdorf]

Attachment: Testcase

Comment 2

[Hixie (not reading bugmail)]

The computed value of 'top' for 'layer2' in the example quoted inline above
should be the distance from the top of 'layer2' when 'position' is 'absolute',
'0' when 'position' is 'relative', and 'auto' if 'position' is 'static' (since
if 'position' is 'static' then there is no computation to complete for 'top').

Comment 3

[Boris Zbarsky [:bzbarsky]]

I'll take this; working on this stuff right now.

Comment 4

[Boris Zbarsky [:bzbarsky]]

OK. I've got some thoughts on a partial fix.  It will not be completely correct.
 To be precise, we will return incorrect values when auto margins/padding are
involved.  Fixing that would require hooks into layout which is bug 106154.

Other than that we should be able to get decent computed values. ETA a few
weeks, I would say.

Comment 5

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

Hooks into layout?  I want to move nsComputedDOMStyle back into layout one of
these days...

Comment 6

[Boris Zbarsky [:bzbarsky]]

That would be awesome.  That's by far the most logical place for it to live
given the sort of things it needs access to.

As it is, I'll just be doing my own calculations for percentages, I guess.  And
fixing logic errors like bug 84599, of course.

Comment 7

[Boris Zbarsky [:bzbarsky]]

Fixed with checkin for bug 84599.  Follow-up bug 109253 filed on coming up with
a spec and implementation of positioning properties for statically positioned
elements.