Closed
Bug 884795
Opened 11 years ago
Closed 11 years ago
DOM built incorrectly for deeply nested <code> elements with the same class name.
Categories
(Core :: DOM: HTML Parser, defect)
Tracking
()
RESOLVED
FIXED
mozilla30
People
(Reporter: sam.scott, Assigned: wchen)
References
Details
Attachments
(1 file)
User Agent: Mozilla/5.0 (Windows NT 6.1) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/27.0.1453.116 Safari/537.36
Steps to reproduce:
DOM is not being built correctly for deeply nested <code> elements with the same class name. Here's a web page that demonstrates the bug - the final text node in the outermost <code> element appears as a child of the <p> element (a sibling of what should be its parent <code> element). If you remove the innermost code elements or change their class names to something else (e.g. "new2") the DOM gets built correctly. This same behavior can be reproduced on Opera, Firefox, Chrome, and IE. But Safari builds the DOM correctly.
<!--
To change this template, choose Tools | Templates
and open the template in the editor.
-->
<!DOCTYPE html>
<html>
<head>
<title></title>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
<style type="text/css">
code.first {
margin-left:20px;
margin-top:10pt;
margin-bottom:10pt;
font-size: inherit;
display:block;
}
code.new {
margin-left:30px;
display:block;
}
</style>
</head>
<body>
<p>The JavaScript code below is from the |ajaxTime.html| example. Take a moment to read it over and try to understand it.
<code class='first'>
$(document).ready( function() {
<code class='new'>
var launchAjax = function() {
<code class='new'>
$.get(
<code class='new'>
"servers/timeServer.php",
<br>{
<code class='new'>
timezone: $("[name=timezone]").val(),
<br>request: $("[name=type]").val(),
<br>delay: $("[name=delay]").val(),
</code>
},
<br>function(data, textStatus, jqXHR) {
<code class='new'>
$("#result").html(data);
</code>
};
</code>
)};
</code>
$("#ajaxButton").click(launchAjax);
</code>
});
</code>
</p>
<script type="text/javascript">
alert(document.querySelector(".first").lastChild.data);
</script>
</body>
</html>
Here's another page that also shows the bug. If you remove the attribute and value from the outermost code element, the bug goes away:
<code bug="true">
a
<code>
b
<code>
c
<code>
d
<code>
e
</code>
f
<code>
g
</code>
h
</code>
i
</code>
j
</code>
k
</code>
|
||
Comment 1•11 years ago
|
||
This is per spec, see the outer loop counter at in the Adoption Agency Algorithm: <http://www.whatwg.org/html/#adoptionAgency>.
Status: UNCONFIRMED → RESOLVED
Closed: 11 years ago
Component: DOM → HTML: Parser
Resolution: --- → INVALID
|
||
Comment 2•11 years ago
|
||
Wait, why is the adoption agency algorithm involved here at all? I thought that only deal with tables stuff?
|
|
||
Comment 3•11 years ago
|
||
Or rather with misnested stuff, but this is not misnested, right?
I think if you're going to mark this as resolved, I need a bit more information. The code was nested properly, so why does the policy for misnested tags apply here?
Status: RESOLVED → UNCONFIRMED
Resolution: INVALID → ---
|
||
Comment 6•11 years ago
|
||
Chrome and IE 10 behaving exactly like Firefox here, so this looks like a spec bug and off like a bug in how Firefox follows the spec. Confirming in the sense that we should follow up by reporting a spec bug and then see how it gets resolved.
Status: UNCONFIRMED → NEW
Ever confirmed: true
Flags: needinfo?(hsivonen)
|
|
||
Updated•11 years ago
|
|
||
Comment 7•11 years ago
|
||
Testcase: http://software.hixie.ch/utilities/js/live-dom-viewer/saved/2446 I really don't understand what's going on here.
|
|
||
Comment 8•11 years ago
|
||
This is related to the Noah's Arc clause: http://www.whatwg.org/html/#push-onto-the-list-of-active-formatting-elements What happens is: <code a> <!-- A, list is A --> <code> <!-- B, list is A,B --> <code> <!-- C, list is A,B,C --> <code> <!-- D, list is A,B,C,D --> <code> <!-- E; B gets removed from the list so list is A,C,D,E --> </code> <!-- closes E, list is A,C,D --> </code> <!-- closes D, lis is A,C --> </code> <!-- closes C, list is just A --> </code> <!-- closes A, because the list takes precedence over the stack --> I think the solution is to make the AAA check the stack first and only do its stuff if the current node isn't a matching formatting element. Filed spec bug: https://www.w3.org/Bugs/Public/show_bug.cgi?id=22926
|
Assignee | |
Updated•11 years ago
|
QA Contact: wchen
|
|
Assignee | |
Updated•11 years ago
|
Assignee: nobody → wchen
QA Contact: wchen
|
|
Assignee | |
Comment 9•11 years ago
|
||
Filed spec bug to make the fix behave a bit better: https://www.w3.org/Bugs/Public/show_bug.cgi?id=24833
|
|
Assignee | |
Comment 10•11 years ago
|
||
|
|
Assignee | |
Comment 11•11 years ago
|
||
Comment on attachment 8386424 [details] [diff] [review] HTML parser AAA algorithm should only pop when the current node is not in the list of active formatting elements. http://www.whatwg.org/specs/web-apps/current-work/multipage/tree-construction.html#adoption-agency-algorithm Implemented the new step 1 of the AAA. Also added some comments and cleaned up existing comments that were strangely formatted.
Attachment #8386424 -
Flags: review?(hsivonen)
|
|
||
Comment 12•11 years ago
|
||
Please do let me know how this goes. I'd be fascinated to hear your implementation experience.
|
|
||
Comment 13•11 years ago
|
||
Comment on attachment 8386424 [details] [diff] [review] HTML parser AAA algorithm should only pop when the current node is not in the list of active formatting elements. s/Arc/Ark/ in the comment. r+ with that. Thanks for figuring this out.
Attachment #8386424 -
Flags: review?(hsivonen) → review+
|
|
Assignee | |
Comment 14•11 years ago
|
||
https://hg.mozilla.org/integration/mozilla-inbound/rev/ac15b6e719be
Flags: in-testsuite+
OS: Windows 7 → All
Hardware: x86 → All
|
||
Comment 15•11 years ago
|
||
https://hg.mozilla.org/mozilla-central/rev/ac15b6e719be
Status: NEW → RESOLVED
Closed: 11 years ago → 11 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla30
You need to log in
before you can comment on or make changes to this bug.
Description
•