53663 - position:absolute table's top/left 'auto' treated as 0 [ABS POS]

Status: RESOLVED FIXED

(Core :: Layout: Tables, defect, P2)

Description

[Jesse Ruderman]

[Split from bug 51037]

Netscape 4.75 and IE both 5.5 both ignore position:absolute when top and left 
aren't given, but Mozilla acts as if both were zero.  Including top:;left:; in 
the style string has the same result.

Moz 2000091808, Windows 98.

Comment 1

[Jesse Ruderman]

Attachment: testcase

Comment 2

[Christine Hoffman]

Attaching essentially the same testcase (just cleaned up HTML - content was in 
the HEAD in previous test).

Is this bug dependent or related ot #10209?

Comment 3

[Christine Hoffman]

Attachment: same testcase with cleaner HTML

Comment 4

[Hixie (not reading bugmail)]

ccing Buster, Karnaze, Pierre; Assigning to Marc; Taking QA; relnote, [ABS POS].

I twiddled the testcase, and the problem also occurs if you explicitly set
the 'top' and 'left' properties to 'auto'. The problem does _not_ occur if you
position, say, a <div>. I'm guessing this is something to do with the messed
up style contexts we have on table elements, but I'm probably totally wrong...

RELEASE NOTE ITEM
   Mozilla currently does not support the value 'auto' on the 'top' and
   'left' properties on tables. It incorrectly treats them as '0' instead. 
   As a work around, authors are encouraged to polute their markup by wrapping
   positioned <table> elements in a <div>, and positioning that, instead.
   For example, instead of: 
      <table style="position:absolute">  ...   </table>
   ...authors could use:
      <div style="position:absolute"><table>  ...   </table></div>
   This should have a similar effect.

Comment 5

[Marc Attinasi]

Accepting. Ian, you are probably correct about the root of the problem, but I'll
have to look closer when we get a chance.

Comment 6

[ekrock's old account (dead)]

Nominating nsbeta1 for goal of achieving cross-browser and backward 
compatibility.

Comment 7

[Bernd]

Attachment: hack

Comment 8

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

That's wrong -- it will break inheritance of 'auto' values and instead cause the
child's value to act like 'auto' rather than inheriting the computed value. 
(Admittedly, there's been some discussion in the working group over whether that
should work, but it does work for us.)

Comment 9

[basic]

status? DBaron: if the patch is wrong, what should be done in this case?

Comment 10

[Greg K.]

Reconfirmed using FizzillaCFM/2002071208. Setting All/All.

Comment 11

[Bernd]

there are two ways to fix the bug IMHO:
- remove the outer table frame and as a consequence the style attributes would
be retrieved for the table itself
- hack around at
http://lxr.mozilla.org/seamonkey/source/layout/html/base/src/nsHTMLReflowState.cpp#1031
 and make there a special treatment if a outer-table frame is encountered.

Comment 12

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

But this bug is only for 'top' and 'left' of auto.  The real problem seems to be
that 'inherit' isn't what you want in this case (for "inherit"-ance from table
to table-outer), since 'inherit' inherits computed values, but you want to
"inherit" the specified value.

Comment 13

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

(But yes, making the code that looks at 'top' and 'left' know about outer-tables
ought to fix the bug.)

Comment 14

[Bernd]

*** Bug 188647 has been marked as a duplicate of this bug. ***

Comment 15

[Bernd]

*** Bug 189257 has been marked as a duplicate of this bug. ***

Comment 16

[Hixie (not reading bugmail)]

Test case:
   http://www.hixie.ch/tests/adhoc/css/box/absolute/008.html

Comment 17

[Boris Zbarsky [:bzbarsky]]

.

Comment 18

[Chris "SoopahMan" Moschini]

Still does not work, Mozilla 1.4.

It sounds like the nature of this problem is that, Mozilla internally thinks of
a single Table Element as 2 elements, an outer frame and an inner, and the inner
dismisses the position settings (top, left, right, etc)?

Trying to get a handle on the C++ code linked to... .

Comment 19

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

Fixed by checkin of bug 205790.

However, Ian's testcase still doesn't work because of a z-ordering bug that
should probably be filed if it isn't already...

Comment 20

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

Actually, never mind about the z-ordering bug.  The testcase requires scrolling...

Comment 21

[Ger Teunis]

*** Bug 220906 has been marked as a duplicate of this bug. ***

Comment 22

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

*** Bug 220906 has been marked as a duplicate of this bug. ***