153118 - Conn: javascript <script src="..."> doesn't timeout

Status: RESOLVED INVALID

(Core :: Networking, defect)

Description

[John Borden]

From Bugzilla Helper:
User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.0; en-US; rv:1.0.0) Gecko/20020530
BuildID:    2002053012

This bug is associated with the Washington Post web site. When I open the page
in Mozilla, everything on the page is displayed except for the text of the
article I'm trying to read. The information about the page indicates it is being
read in quirks mode. The source for the page actually does contain the missing
text. The problem is the same for all articles that I tried to read. Internet
Explorer does not exhibit this problem. I don't have a copy of Netscape to try.
I tried changing just about every option in Mozilla with no effect. 

Reproducible: Always
Steps to Reproduce:
1.open the page
http://www.washingtonpost.com/wp-dyn/articles/A13114-2002Jun19.html?referer=email
2.
3.

Actual Results:  Page displayed without the text of the article. Message at the
bottom of the screen says "Transferring data from Washingtonpost.com". Sometimes
it says "Transferring data from ad.doubleclick.com". Ok, now that I've done it a
few time, it mostly says "Transferring data from ad.doubleclick.com". 

Expected Results:  Display the text

Comment 1

[Oddsson]

I can verify this using build 2002061908 on Win2k.

View->Source shows that the HTML has loaded completely (</html> at the end) but
the hourglass mouse cursor is still shown and Moz still says it's transferring
data from washingtonpost.com (I have Mozilla block all Doubleclick ads).

Comment 2

[Boris Zbarsky [:bzbarsky]]

to parser; I'm not seeing the text in the content model.

Comment 3

[Paul Bergsagel]

I have the text display using Mac OS X and Mozilla 1.1a.

Comment 4

[[e-mail]]

I can confirm on WinXP, build 2002070608. The page loads forever and doesn't
display the article.

When javascript is disabled, the page loads and displays fine.

And what is really strange: When I disable javascript, and then enable
javascript, the page loads and displays fine again. I need to restart Mozilla
(with javascript enabled) to get the bug again.

Comment 5

[[e-mail]]

Attachment: HTML source, somewhat simplified

I tried to create a test case and made the following observations:

* when the page rendered correctly one time (due to me having deleted too
much), a restart of Mozilla (sometimes even of the computer) is needed, to
reliably reproduce the error again

* the bug disappears eratically, when making minor (and seamingly unrelated)
changes to the source, like deleting or inserting whitespace

* it took me some hours to partially simplify this case, I would be glad if
someone could simplify further or confirm that the error still exists in the
simplified version

Comment 6

[[e-mail]]

Attachment: good testcase

Ok, I got this one.

There is a javascript directive that document.writes a script tag that refers
to some script URL. Upon loading the script from that URL, Mozilla *does not
timeout*. If the server doesn't send any data, Mozilla simply sits there and
waits forever and is not displaying any HTML code that is below the script tag.
I'd say this is a networking bug.

The testcase needs to be edited to get it to work. It needs an open port on
some server, that just accepts connections but does nothing else. This can be
done by using netcat, e.g.:
  # nc -l -p 2000
will create a process that just listens on port 2000.

This is always reproducible, btw.

Comment 7

[harishd]

Yup, that smells like netlib bug.

--> Darin

Comment 8

[[e-mail]]

On the other hand, this might be a feature, not a bug.

IE6 behaves the same way.

And when I try to load a page from a server that doesn't send anything, just
accepts the connection (same netcat thing), Mozilla doesn't time out either.

Summary: I have the impression, that for HTTP connections in general, as long as
the TCP connection is established alright and not being torn down, Mozilla will
not ever time out. I personally don't like this very much, but I'd guess that it
is a feature, not a bug. Maybe HTTP protocol spec require this.

Comment 9

[[e-mail]]

Attachment: better testcase

I made the testcase a little smaller, it still needs to be edited to work.

Comment 10

[[e-mail]]

Bug 70081 might be similar or duplicate but it is too old to verify that.

Comment 11

[benc]

UPDATE:
mozilla 1.1 - the testcase seems to work now...

mozilla does have an HTTP application level time-out, but I think the question
for the situations you described was: is there no connection, or a very slow
connection that never completes.

Comment 12

[John Borden]

Based on Benjamin's comment, I installed 1.1 on a different computer and checked
for the bug. It did not appear, neither on the original web page nor on either
of the test cases. Then I went back to the other computer that still has 1.0 on
it and tried it there to see if it was still happening. The bug did not appear
on the original web page or either test case. Strange... 

So, I guess the bug is fixed, but it does not appear that it was any changes to
Mozilla that fixed it. Since first reporting this, I have installed OS updates
and a new Java Runtime. I would have to guess that it was one of those that
fixed it.

Oh, BTW, Thiemo suggested that IE6 behaves the same way. When I first reported
the bug I was also running IE6 and it did not exhibit the bug under identical
circumstances. 

I leave it someone else to decide if this bug should be marked as resolved. I
don't like it when things start working for unexplained reasons.

Comment 13

[benc]

phil: can you lend any javascript expertise to this problem?

Comment 14

[Phil Schwartau]

ben: sorry, I can't on this one. The JS Engine only takes over
after a script has been successfully loaded -

Comment 15

[Samir Gehani]

adt: nsbeta1-

Comment 16

[Darin Fisher]

-> default owner