51230 - Width of Absolutely positioned, non-replaced elements calculated incorrectly.

Status: RESOLVED INVALID

(Core :: Layout, defect, P3)

Description

[Val Sharp]

From Bugzilla Helper:
User-Agent: Mozilla/4.6 [en-gb]C-CCK-MCD NetscapeOnline.co.uk  (Win98; I)
BuildID:    2000080712

According to CSS2 10.3.7, the Width of an Absolutely positioned non-replaced 
element, plus left, right, margins, borders and padding, should equal the width 
of the containing block.

This is similar to the calculation of Width for Block-level, non-replaced 
elements in normal flow - for our purposes, if left, right and width are default 
auto, margins are default zero, and border-style is default none, then the 
calculated width should also be the same.

Going through the steps in 10.3.7:
  (Border width => 0)
1. Left => 0
2. n/a
3. Right=> 0
4. margins already 0 (default)
5. n/a
6. solve equation for only remaining auto 
   i.e. width => width of containing block
7. Right remains 0

However, in M17 the Absolutely positioned element is instead just wide enough 
for its contents.


Reproducible: Always
Steps to Reproduce:
The attachment consists of four DIVs: 
- two Block-level, non-replaced elements in normal flow - one static and one 
relative, and
- two Absolutely positioned, non-replaced elements - one fixed and one absolute.


Actual Results:  Notice that the widths are calculated correctly for the first 
two DIVs, but incorrectly for the second two.
			

Expected Results:  In this example, Widths of Absolutely positioned DIVs should 
equal width of containing block.

Comment 1

[Val Sharp]

Attachment: Auto Width testcase

Comment 2

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

There are (still unpublished, for some reason) errata to the CSS2 spec changing
these rules.

Comment 3

[Hixie (not reading bugmail)]

Taking bug. I will investigate this at some point. However, I believe we are
probably following the [new] rules correctly.

Comment 4

[Keyser Sose]

Marking NEW since bug is under investigation.

Comment 5

[Hixie (not reading bugmail)]

Upon managerial request, adding the "testcase" keyword to 84 open layout bugs that
do not have the "testcase" keyword and yet have an attachement with the word
"test" in the description field. Apologies for any mistakes.

Comment 6

[Val Sharp]

'Errata in REC-CSS2-19980512' - 
http://www.w3.org/Style/css2-updates/REC-CSS2-19980512-errata.html - 
was published 2000/10/24. It contained Changes to the calculation of the 
_height_ of abs pos non-replaced elements, but not the width.

Does anyone have any knowledge of future changes to the _width_ calculation? 
Otherwise, this is a bug which should be fixed.

Comment 7

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

Near the very bottom of the page you site there is a "[Text forthcoming]" for
the errata for widths.  The changes should be similar to the ones for heights.

Comment 8

[Val Sharp]

OK. Hadn't noticed publication of later version of Errata. Thanks.

Comment 9

[Hixie (not reading bugmail)]

Does that mean this bug is INVALID after all?

Comment 10

[Val Sharp]

Maybe it means there is a very high probability that this bug _will be_ INVALID.

Pity it can't be RESOLVED 'PENDING'...

Comment 11

[karnaze (gone)]

moving to m1.0.1.

Comment 12

[Jason Johnston]

It is my understanding that "absolutely positioned non-replaced elements" in the
CSS2 spec actually refers to position:relative -- position:absolute and
position:fixed are "absolutely positioned replaced elements".

If this interpretation is correct, then Val's testcase exhibits the correct
width calculation for all position types: position:static and position:relative
calculate widths with reference to the containing block, and position:absolute
and position:fixed calculate width according to the intrinsic width of their
contents.
(tested on Win2K build 2001072003)

Can someone confirm this interpretation?

Comment 13

[Hixie (not reading bugmail)]

'absolutely positioned elements' are those with position:absolute;
'fixed positioned elements' are those with position:fixed;
'relatively positioned elements' are those with position:relative;
'non-positioned elements' are those with position:static;
'replaced elements' are images, applets, IFRAMEs and the like;
'non-replaced elements' are everything else.

Thus an 'absolutely positioned replaced element' is an image with position:
absolute, which an 'absolutely positioned non-replaced element' is any other
element with position:absolute.

Comment 14

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

...although 'absolutely positioned elements' can in some cases refer both to
those with 'position: absolute' and those with 'position: fixed'

Comment 15

[Greg K.]

Reconfirmed using FizzillaCFM/2002071208, at least as originally reported; boxes
3 and 4 are only as wide as their contents. Setting All/All.

Comment 16

[Charles Goodwin]

Is this behaviour of mozilla on this issue currently acceptable?

From 10.3.7 of the css2 spec:

"The constraint that determines the computed values for these elements is:

    'left' + 'margin-left' + 'border-left-width' + 'padding-left' + 'width' +
'padding-right' + 'border-right-width' + 'margin-right' + 'right' = width of
containing block "

Rearranging this, you can say:

'width' = width of containing block - [all other variables]

Yet, in Mozilla, if you specify a style for a top level <div> (ie containing
block = page) as such:

div.testcase {
    left: 50px;
    right: 50px;
    width: 100%;
}

The width is evaluated as the width of the page by Mozilla.  Surely this is
incorrect, as it should be [page width - (left + right)].  This is how IE, at
least, evaluates it.  And the differing behaviour kinda impedes cross-browser
css.  (Not that that is the reason it should be changed.  It should be changed
because it is incorrect.)

Comment 17

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

1) Please read CSS 2.1, not CSS 2, for positioning rules.
2) Percentages on 'width' *always* refer to the containing block width, period.

Comment 18

[Boris Zbarsky [:bzbarsky]]

http://www.w3.org/TR/CSS21/visudet.html#abs-non-replaced-width (CSS 2.1, section
10.3.7) has the new positioning rules.  Going through them:

pre-numbered-list:  "If all three of 'left', 'width', and 'right' are 'auto': if
                     'direction' is 'ltr' set 'left' to the static position and
                     apply rule number three below"

Rule #3 is: "'width' and 'right' are 'auto' and 'left' is not 'auto', then the
             width is shrink-to-fit . Then solve for 'right'"

Which is what we're doing.