77072 - page fails to load, with image not displayed

Status: RESOLVED WORKSFORME

(Core :: Graphics: ImageLib, defect, P3)

Description

[Jesse Ruderman]

Setup:
1. If you're using Windows, set up your mozilla.exe shortcut to use the 
-console option.
2. Put the "linked images" bookmarklet on your personal toolbar.
   a. Go to http://www.squarefree.com/bookmarklets/pagelinks.html#linked_images
   b. Right-click on the "linked images" link and select "add to bookmarks".
   c. Put the bookmarklet in your personal toolbar folder.

Steps to reproduce:
3. Restart Mozilla (optional).
4. Open one or two browser windows to http://dmoz.org/images/moz/.
5. Use the bookmarklet in one window.  This will open a window showing all of 
the images on http://dmoz.org/images/moz/.
6. Wait for the page with the images to load completely.
7. Close the page with the images.
8. In the any window with http://dmoz.org/images/moz/ loaded (not necessarily 
the same one as in step 4), use the bookmarklet again.

Result:
* Stop button disappears except when hovered over.
* Throbber becomes blank and dark blue (Modern2) or grey (Classic).
* Links stop working in all windows until I restart Mozilla.  I still get a 
hand cursor when I hover over links, but I no longer see status bar text for 
links.  Bookmarks still work.
* Opening new windows stops working (Ctrl+N, run mozilla.exe again) until I 
restart Mozilla.
* Closing all browser windows doesn't exit Mozilla.  Have to close the console 
window or use Ctrl+Alt+Del to kill Mozilla before it can be restarted.

Reproducibility: about 75%.

Comment 1

[benc]

mass move, v2.
qa to me.

Comment 2

[Bradley Baetz (:bbaetz)]

-> darin, after discussion with him.

netstat is showing we're leaking sockets. This is probably the bug darin earlier
when we stress the socket thread, but couldn't reproduce reliably.

Comment 3

[Jesse Ruderman]

All/all, since bbaetz was able to reproduce this on linux.

Comment 4

[Darin Fisher]

a much simpler testcase are the following links from
http://cowtools.mcom.com/cgi-bin/page-loader/loader.pl:

www.tomshardware.com  (leaves XXX sockets w/ 1 byte in Recv-Q)
www.voodooextreme.com (leaves two sockets w/ 1 byte in Recv-Q)

running over a 56k connection (eg. shaper0 @ 56600 on cowtools).

Comment 5

[Darin Fisher]

www.tomshardware.com (leaves two sockets w/ 1 byte in Recv-Q)

Comment 6

[Darin Fisher]

much of the problems reported in this bug have (i think) been fixed by my patch
for bug 82241.  i believe now we are just left with the problem that the page
doesn't finish loading.  the testcases on cowtools (www.voodooextreme.com and
www.tomshardware.com) behave similarly (i think).

this bug no longer blocks bug 86274.  removing the hang keyword.

Comment 7

[Darin Fisher]

*** Bug 85357 has been marked as a duplicate of this bug. ***

Comment 8

[Darin Fisher]

reducing priority since this shouldn't make the mozilla session useless anymore.

Comment 9

[Darin Fisher]

-> pavlov, since necko seems to have finished downloading everything.

Comment 10

[Darin Fisher]

once more...

Comment 11

[Stuart Parmenter]

-> 0.9.3

Comment 12

[benc]

->imagelib

Comment 13

[Bradley Baetz (:bbaetz)]

*** Bug 91501 has been marked as a duplicate of this bug. ***

Comment 14

[Bradley Baetz (:bbaetz)]

Changing summary.

See http://www.defleppard.com/defleppard/defleppard.html for a reproducable test
case.

Comment 15

[keaeris]

I attempted to reproduce the sequence here this morning on my Mac(9.1).  In my
case, I didn't have to use the bookmarklet to get the behavior and nothing
changed after I did so. Flushing the memory and disk caches does not correct the
result, but different images may fail to load after a reload.

Comment 16

[Jason Kersey]

-> 0.9.4

Comment 17

[Pierre Saslawsky]

It maybe related to bug 82720 and bug 92354.

Comment 18

[Stuart Parmenter]

.

Comment 19

[Jesse Ruderman]

Regression of the original bug, or at least very similar: bug 104769, "Some JS 
created windows become invisible zombies".

Comment 20

[scottputterman]

Moving Netscape owned 0.9.9 and 1.0 bugs that don't have an nsbeta1, nsbeta1+,
topembed, topembed+, Mozilla0.9.9+ or Mozilla1.0+ keyword.  Please send any
questions or feedback about this to [e-mail].  You can search for
"Moving bugs not scheduled for a project" to quickly delete this bugmail.

Comment 21

[Gagan]

per adt, not critical for nsbeta1. hence minus.

Comment 22

[Matthias Versen [:Matti]]

Jesse: is this still a problem with recent builds ?

Comment 23

[Jesse Ruderman]

WFM in Seamonkey 07/04.