265814 - The output of document.write() is always (first) parsed in quirks mode

Status: RESOLVED FIXED

(Core :: DOM: HTML Parser, defect)

Description

[Blake Kaplan (:mrbkap) (inactive)]

I came across this when investigating bug 165323. Basically, when we come across
a document.write() in the body, we try to parse just the contents of that
write() as their own block. That parse is always performed in quirks mode in the
parser. If the parse fails for any reason (such as an unclosed comment) the data
is then inserted into the regular input stream, and parsed in standards mode.
I'll attach a testcase shortly.

Comment 1

[Blake Kaplan (:mrbkap) (inactive)]

Attachment: testcase

Comment 2

[Blake Kaplan (:mrbkap) (inactive)]

Attachment: better testcase

The last testcase was sort of unclear (it showed the problem, but in a
roundabout way). This one is much more self-explanatory.

Comment 3

[Blake Kaplan (:mrbkap) (inactive)]

Attachment: patch v1

This patch ensures that we bring over the DTD mode from the last context (since
nsHTMLDocument::WriteCommon() doesn't send one in). I'm not sure if we can get
here without a mPrevContext for the new parsercontect, so I'm not sure if the
else clause is strictly necessary, but I thought it wouldn't hurt to be
defensive.

Comment 4

[Blake Kaplan (:mrbkap) (inactive)]

Comment on attachment 163184 [details] [diff] [review]
patch v1

bz, since you know about this, I'll ping you for r (and maybe sr?)=.

Comment 5

[Boris Zbarsky [:bzbarsky]]

Comment on attachment 163184 [details] [diff] [review]
patch v1

r=bzbarsky if you only do the preservation thing when |aMode ==
eDTDMode_autodetect|.

jst, would you sr?

Comment 6

[Blake Kaplan (:mrbkap) (inactive)]

Attachment: patch v2

Comment 7

[Johnny Stenback (:jst)]

Comment on attachment 163184 [details] [diff] [review]
patch v1

sr=jst

Comment 8

[Blake Kaplan (:mrbkap) (inactive)]

Checked in.