24184 - Parser doesn't close <span> in residual-style mode

Status: VERIFIED FIXED

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

Description

[jmraker]

Using Win32 2000011708 Build.

In this testcase, the CSS properties appear outside of their scope.
There are various ways of making it go away, like: Adding a space at the end of
a <P>, removing the global <font> tag, and removing the first <p>...</p>
paragraph.  This bug does not appear in IE, or Netscape 4.7

Expected Results:
All 3 <hr> seperated sections should look alike.

Actual Results:
Due to a font-size: of over 100%, each section gets larger and larger and has
properties of previous CSS tags.

Comment 1

[jmraker]

Attachment: testcase for bug

Comment 2

[Hixie (not reading bugmail)]

Attachment: Slightly clearer test case

Comment 3

[Hixie (not reading bugmail)]

Rick, this is definitely a strange parser problem!

It requires a <font> tag, a <p> tag, and a comment with no surrounding
whitespace, as well as some even more contrived conditions for where those
elements go!!!

jmraker: Very nice catch!

Comment 4

[harishd]

This ought to be a layout bug.  Here is the content-model [part of ]:

  body@011FC52C refcount=3<
    Text@01210390 refcount=3<\n>
    font@0121066C face=Arial refcount=3<
      Comment@01211700 refcount=2<!-- Deleting this font tag fixes the bug -->
      Text@01211DE0 refcount=3<\n>
    >
    p@0121177C refcount=3<
      font@0121392C face=Arial refcount=3<
        Text@01213F40 refcount=3<testcase 4>
      >
    >
Here is the content-model for the FONT tag [ in question ] ended by /FONT. That 
is, "<FONT FACE="Arial"><!-- Deleting this font tag fixes the bug -->
</FONT><P>...":

  body@011F5DAC refcount=7<
    Text@0120F750 refcount=3<\n>
    font@0120F95C face=Arial refcount=3<
      Comment@01210970 refcount=2<!-- Deleting this font tag fixes the bug -->
      Text@01210B70 refcount=3<\n>
    >
    p@01210C4C refcount=3<
      Text@01210CB0 refcount=3<testcase 4>
    >

Note that the models are identical.  

Comment 5

[harishd]

Attachment: attaching test case that works..

Comment 6

[harishd]

Ooops, I take my comments back.. I did not look into the content-model properly.
I do the see the problem now!!

Comment 7

[rickg]

Harish -- I don't see the problem. Can you be more specific?

Comment 8

[harishd]

The bug seems to have gone away!!!

Here is the current content-model that is identical to the second model shown
in comment 000-01-24 10:19:

  body@0203AADC refcount=3<
    Text@0206B7A0 refcount=3<\n>
    font@0206B73C face=Arial refcount=3<
      Comment@0206FCE0 refcount=2<!-- Deleting this font tag fixes the bug -->
      Text@0206FC70 refcount=3<\n>
    >
    p@0206FA6C refcount=3<
      Text@0203CF00 refcount=3<testcase 4>
    >

Note: The above content-model is based on test case ( id=4305 ).

Marking FIXED.

Comment 9

[jmraker]

On Win32 2000022716 daily build:

The bug never went away (I didn't notice it either until last week).  The <HR>
(in testcase 1) were indenting (they shouldn't).
Today I look again, and the parser bug is now affecting every class=a and the
first class=z in testcase 2.

Comment 10

[Mark]

Not closing the <SPAN> tag makes ZDNET articles look bad:
http://www.zdnet.com/zdnn/stories/news/0,4586,2523934,00.html

Comment 11

[Michael Kay]

This is causing problems on HotBot.com.
See: http://staff.hotwired.com/mikeque/mozilla/red1.html

Comment 12

[harishd]

Attachment: Try this patch. ( I will check this in over the weekend ).

Comment 13

[harishd]

Attachment: Ignore the previous patch....try this.

Comment 14

[harishd]

Tweaked DTD's containment logic. 

Attachments in comment 2000-04-09 17:46 and 2000-04-12 17:29 are slightly 
different from this bug. Will open a separate bug. 

Marking this bug FIXED. 

Comment 15

[jmraker]

Using Win32 Nightly build 200050208:

I still see this bug in every nightly build.  Please re-evaluate the status of 
this bug if the fix that closed this is in the nightly builds.

Comment 16

[harishd]

I see this too...mmm...wonder what happened. Reopening the bug.

Comment 17

[harishd]

Not a beta stopper... moving to M17.

Comment 18

[jmraker]

With the Linux nightly 2000050908 build:

The testcase now renders correctly.

Comment 19

[harishd]

Renders correctly on NT too. Marking FIXED.

Comment 20

[bsharma]

updated qa contact.

Comment 21

[bsharma]

Verified on
build: 2001-05-29-20-Trunk
platform: Win NT

The test case works fine.