153753 - HTML STYLE elements should be closed by first </ sequence

Status: VERIFIED WONTFIX

(Core :: DOM: HTML Parser, defect)

Description

[Greg K.]

According to the HTML 4.0.1 specification, Section 6.2, the STYLE element should
be closed by the first encountered </ sequence, not merely the first encountered
</STYLE> sequence. Only in valid documents will the latter sequence be encountered.

Comment 1

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

HTMLParser bug.  We might want to consider fixing for strict mode (perhaps even
for quirks mode for stylesheets, but it would definitely break scripts).

Comment 2

[Greg K.]

HTML 4 reference: [http://www.w3.org/TR/html401/types.html#h-6.2].

Comment 3

[Greg K.]

Attachment: STYLE element testcase

A quick testcase for the STYLE element. Note the first statement (on 'div')
following the </foo> is ignored, as expected, but the following statement (on
'p') is not ignored.

Comment 4

[Greg K.]

Attachment: SCRIPT element testcase

A quick testcase of the SCRIPT element. It generates a syntax error on the JS
Console, so the effective behavior is as described. Should the parser simply
interpret the </foo> as the end of the script, though, and not send it to the
JS engine?

Comment 5

[Boris Zbarsky [:bzbarsky]]

<script>
document.write('<a>text</a>');
</script>

is very common and is why we do not do this per spec at the moment... Doing this
for scripts would break a large fraction of pages out there....

I agree that we should do this for <style>, though....

Comment 6

[Frankie]

5 months later...

First, this bug should be platform All/All rather than Mac/OSX. I tested Win98
release 1.2.1 and got the same results.

Second, the testcases should at least TRY to validate. There are ways to place a
</ sequence in a style or script that would otherwise be valid, rather than
throwing an arbitary </foo> in the middle:

script testcase: document.writeln("<i>foo</i>")
style testcase: div:before { content: "<i>foo</i>" }

Third, I think the script side of this bug should be WONTFIX, because it would
break way too many websites.

Comment 7

[Greg K.]

Should the SCRIPT part be implemented in Standards-compliance mode only, then,
as mentioned in comment 1?

Comment 8

[Boris Zbarsky [:bzbarsky]]

If you're willing to fork some of the most fragile code in the app and create a
variant that will receive almost no testing and hence will break often, sure it
could be done.  Just a matter of time and priorities.  ;)

Comment 9

[Blake Kaplan (:mrbkap) (inactive)]

Boris, this actually shouldn't be too hard to implement.

Is </ common in <style> tags? While the testcase in comment 6 is points to a
quirks-only fix, I wonder if we should just follow the spec in all cases.

Should I bother (and if I should, should I wait until an alpha cycle)?

Comment 10

[Boris Zbarsky [:bzbarsky]]

See comment 1.  Fixing this for <style> may be OK, but fixing it for <script>
will absolutely break pages; I'm against doing it in any mode.

The only way <style> will contain a "</" is if it's present in one of the
strings in the stylesheet...  Probably rare, since strings in stylesheets are
somewhat rare.  But still it may be worth wontfixing this on the basic premise
that absolutely no UA implements this part of HTML (assuming that's true).

Comment 11

[Blake Kaplan (:mrbkap) (inactive)]

Since no other browser implements the spec. in this way, this bug is WONTFIX.
There is a message
(http://listserver.dreamhost.com/pipermail/whatwg-whatwg.org/2005-January/002993.html)
on the WHATWG mailing list calling this part of the HTML spec. into question,
and (even in standards mode) this fix would only cause web sites grief.

Comment 12

[Blake Kaplan (:mrbkap) (inactive)]

*** Bug 150323 has been marked as a duplicate of this bug. ***