615868 - images not placed properly with css

Status: RESOLVED WONTFIX

(Core :: DOM: HTML Parser, defect)

Description

[Olivier LEVILLAIN]

User-Agent:       Mozilla/5.0 (Windows NT 5.1; rv:2.0b7) Gecko/20100101 Firefox/4.0b7
Build Identifier: Mozilla/5.0 (Windows NT 5.1; rv:2.0b7) Gecko/20100101 Firefox/4.0b7

When placing images using css + javascript, images are not well placed.
The same test works correctly in FF3.6.12 and in IE8.

Reproducible: Always

Steps to Reproduce:
1. Load the attached earth.html if FF4.0b7
2. Put your mouse over south america
Actual Results:  
Displays north america in blue over south america

Expected Results:  
Should have displayed South America in blue

Another test: place your mouse above south *africa*, it shows Russian logo instead of SA logo

All placements are made through constant x/y data in the earth.css file
Exactly the same test works correctly in FF3.6.12 and in IE8.

Comment 1

[Olivier LEVILLAIN]

Attachment: zip file containing the test case files (html, css and gif)

Comment 2

[(mostly gone) XtC4UaLL [:xtc4uall]]

I see the same with Mozilla/5.0 (Windows NT 5.1; rv:2.0b8pre) Gecko/20101130 Firefox/4.0b8pre = Google Chrome 9.0.597.0.

And MSIE 8 = Opera 11.1104 = FX 3.6.x.

not sure what is correct though.

Comment 3

[dolphinling]

The main problem is code like <a href=""/>

<a> is not an empty element, so the self-closing form isn't valid in html. The error handling mechanism involves duplicating the <a>, so it's in two places on the page, hence the visible result.

Testcase:
data:text/html,<ul><li><a/></li> </ul>

NOT testcase (for comparison):
data:text/html,<ul><li><a/></li></ul>

(note the missing space)

I'm willing to bet firefox's behavior is correct, but can't point to exactly where in the spec it says so.

Comment 4

[Robert Longson [:longsonr]]

This is XHTML being served as text/html, this is not valid because it
violates Appendix C of the XHTML specification.

Comment 5

[Boris Zbarsky [:bzbarsky]]

Yep.  With the old parser, the </li> closed the unclosed <a>.  But now it stays open until it sees the following <a>.

Henri, is that the right behavior?  If so, given that no other browser does that, would this be worth a spec change?

Real testcase that shows that difference:

  data:text/html,<ul><li><a/></li> <li><a/></li></ul>

Robert, just because the content is invalid doesn't mean changing how we render it is necessarily ok!  Most (all, to a first approximation) content on the web is invalid.

Comment 6

[Olivier LEVILLAIN]

(In reply to comment #5)
> Robert, just because the content is invalid doesn't mean changing how we render
> it is necessarily ok!  Most (all, to a first approximation) content on the web
> is invalid.

I totally agree with this sentence! if a simple <a/> in an html page breaks this page, how many millions of pages will be broken in the world? Moreover, if it is actually working in FF3.6 and all other browsers...

Comment 7

[Henri Sivonen (:hsivonen)]

(In reply to comment #5)
> Henri, is that the right behavior?

Per spec, yes.

> If so, given that no other browser does that, would this be worth a spec change?

I think this isn't worth a spec change.

Chrome 9 does exactly what Firefox 4 does. Also, pre-HTML5 WebKit (as seen in Safari and Midori) behaves similarly even if not exactly the same. IE9 PP7 Standards Mode clones pre-HTML5 WebKit here and works exactly like Safari and Midori.

Comment 8

[Boris Zbarsky [:bzbarsky]]

> if a simple <a/> in an html page breaks this page, how many millions of pages
> will be broken in the world?

So far, only one, after many months of testing by hundreds of thousands of people.  ;)

> Moreover, if it is actually working in FF3.6 and all other browsers...

Right; that's the important part.  Except it's not true, apparently, since it doesn't work in Safari and other webkit-based browsers (e.g. in the attached testcase hovering over Egypt in Safari highlights Eastern Europe).

Given that, I agree with the wontfix.  Using valid markup would make this work, for sure.  ;)

Comment 9

[Olivier LEVILLAIN]

(In reply to comment #8)
Ok, I give up :)