47827 - "Stair-stepped" display of stories on page from Left to right.

Status: VERIFIED FIXED

(Core :: DOM: HTML Parser, defect, P3)

Description

[Joey Sum]

From Bugzilla Helper:
User-Agent: Mozilla/5.0 (Windows; U; Win98; en-US; m18) Gecko/20000806
BuildID:    2000080608

On the topic pages of the NY Post website (such as News, Sports, etc.), the
stories are displayed in a "stair-stepped" fashion, as if you were looking at a
flight of stairs from the side, descending from left to right.

Reproducible: Always
Steps to Reproduce:
1.Visit http://www.nypost.com/news/news.htm
2.or visit http://www.nypost.com/sports/sports.shtml
3. or visit http://www.nypost.com/business/business.htm

Actual Results:  I observed the stair-step fashion of the stories in the table.

Expected Results:  Should have rendered the stories in a table with equal
widths. Netscape v4.7x (Win32/Linux) and MSIE v4.x/5.x display these pages as
expected.

Comment 1

[Mats Palmgren (inactive)]

*** Bug 47828 has been marked as a duplicate of this bug. ***

Comment 2

[Mats Palmgren (inactive)]

Attachment: Testcase

Comment 3

[Mats Palmgren (inactive)]

Confirming bug, over to Layout.

Comment 4

[Mats Palmgren (inactive)]

I have analyzed the problem further and it seems that illegal tags (in the URL
<CR>) are affecting layout/parsing. The following two lists are laid out
differently:
<DL>
  <DD>text<BADTAG>
  <DD>text<BADTAG>
  <DD>text<BADTAG>
</DL>

<DL>
  <DD>text
  <DD>text
  <DD>text
</DL>

Comment 5

[Mats Palmgren (inactive)]

Attachment: Testcase #2 (the <BADTAG> example)

Comment 6

[Hixie (not reading bugmail)]

Marc, Pierre, Chris, Eric, Harish: This is another example of the problem with
including unknown tags in our content model for the purposes of roundtripping.

Because we see the <BADTAG> we can't automatically close the <DT>. So each <DT>
becomes nested in the previous one.

Our behaviour is technically valid, since invalid markup == undefined behaviour,
and is even in the spirit of XML...

Maybe we should just assume that any unknown tag is an empty element with no end
tag? (like <br>) Then treat the unexpected end tags just like we do now for
round tripping purposes (however that is).

David Baron believes this is a dup...

Comment 7

[Pierre Saslawsky]

Reassigned to Harish to decide how to close unknown tags.

This is not a dup. Bug 22260 should be fixed for unknown elements that are 
correctly closed.

Comment 8

[harishd]

Attachment: Proposed patch [ to prevent DD nesting ]

Comment 9

[Hixie (not reading bugmail)]

harishd: I take it that the following will still work with your patch, right?:

   <dl>
     <dt> term </dt>
     <dd> 
       <dl>
         <dt> term </dt>
         <dd> I should be nested 2 deep </dd>
       </dl>
     </dd>
   </dl>

Comment 10

[harishd]

Attachment: Scree shot for Ian's previous comment [ looks identical to Nav4.x ]

Comment 11

[harishd]

Content model for Ian's test case ( with my patch ):

docshell=01C6E790
html@0302E4E8 refcount=3<
  head@0302E3F8 refcount=2<
  >
  Text@03077C40 refcount=3<\n   >
  body@03074E78 refcount=3<
    dl@03003AB8 refcount=3<
      dt@03003328 refcount=5<
        Text@030032B0 refcount=3< term >
      >
      dd@03005F88 refcount=3<
        Text@03005F10 refcount=3< \n       >
      >
      dl@03005578 refcount=3<
        dt@03005518 refcount=3<
          Text@030054A0 refcount=3< term >
        >
        dd@03006498 refcount=3<
          Text@03006420 refcount=7< I should be nested 2 deep >
        >
      >
    >
    Text@03007750 refcount=3<\n>
  >
>

It looks correct to me.

Comment 12

[Hixie (not reading bugmail)]

harishd: ok, cool

Comment 13

[Nisheeth Ranjan]

Marking nsbeta3+...

Comment 14

[harishd]

Patch checked in. Marking FIXED.

Comment 15

[harishd]

Marking FIXED....once again!

Comment 16

[Chris Petersen]

Looks fine in the Sept 7th build. Marking verified.

Comment 17

[Andreas Franke (gone)]

There is still a (minor) layout difference for first testcase: 
The "stair-stepped" display (the major problem) has been fixed, but the
indentation of the "By..." lines still differs between Netscape 4.61 and Mozilla
builds from 9/22 and 10/22 (both branch and trunk).

Is this a bug or a feature? Is this tracked elsewhere?

The second testcase is fine.