Closed GoogleCodeExporter closed 8 years ago
Connections on IE6 are still open after the window close event.
This old browser can only handle 2 connections to a web serveur.
This issue is critical ...
See attached patch.
Thx for this great library!
Original comment by pierrick.hymbert
on 6 Apr 2011 at 3:53
Attachments:
im not talking about ie6 specially... i can reproduce this in any browser
when u use the tool fiddler2 it will monitor each connection on any browser
except connections to the localhost
Original comment by Peter.Nu...@googlemail.com
on 6 Apr 2011 at 6:02
The patch fix the issue for all browsers ;-)
Original comment by pierrick.hymbert
on 6 Apr 2011 at 6:52
Hi,
I just tried it on localhost with the latest trunk version (without the patch).
The connections are shown with Firebug and Chrome Developer Tools, but each
connection is closed correctly. Is it not reproducable on localhost? Could you
please try if the problem does also occur with the demo conversation app (war
file attached to the post)?
Thank you in advance and thank you for the patch!
Original comment by sven.strohschein@googlemail.com
on 6 Apr 2011 at 10:14
Attachments:
close the issue...
im sorry i found the problem in front of the monitor :(
Original comment by Peter.Nu...@googlemail.com
on 7 Apr 2011 at 7:17
The problem is that the unlisten() method does not force the end of the user
pending listen() connection.
The listen() method waits until the max.time.wait occurs.
If the browser is closed before the max.time.wait occurs (20s), the connections
are not closed properly.
With IE6 the TCP connections are still CLOSE_WAIT for ever when Ajax request
are not close properly.
@see
http://bytes.com/topic/javascript/answers/643080-ajax-crashes-ie-close-window
http://www.coderanch.com/t/117945/HTML-JavaScript/Too-many-CLOSE-WAIT-due
To reproduce :
1. Add Listeners
2. Remove Listeners by RemoteEventService.removeListeners
3. Monitor connections by Firebug
Thx for your reactivity!
Original comment by pierrick.hymbert
on 7 Apr 2011 at 7:35
I can still not reproduce this issue (also tested with IE9). I could imagine
that the connection is not closed correctly when the max. waiting time is still
active (like perrick described), but this should not happen because an unlisten
event is send when the unlisten()/removeListeners() method is called. This
unlisten event should terminate the waiting connection automatically and that
worked in my tests. Therefore the unlisten() method forces the end of the
listen() connection. Can someone still reproduce this issue?
Original comment by sven.strohschein@googlemail.com
on 3 Oct 2011 at 10:22
We can reliably reproduce this issue. We are using version 1.1.1.
The client sends a request to the gwteventservice to listen for events and the
server responds in 30s (proven by our apache server logs and timestamps),
however, on occasion, the response does not reach the browser. As a result, in
the browser (via HttpFox plugin) it looks like there is a hung connection.
After a period of time, some node on the network sends the brower a TCP/IP FIN
command and only then the browser closes the connection. The issue of the
brower not receiving the response even though the server has sent a response is
clearly a network issue. However, the client should be more robust in how it
handles bad networks. Is there anything that can be done within the
gwteventservice javascript code that can deal with long connections that
surpass the configured timeouts?
thanks,
JK
Original comment by j...@architech.ca
on 1 Nov 2011 at 6:05
I could still not reproduce the issue in a live test, but the listening request
is now terminated/returned directly when the user gets removed (via unlistening
/ removing listeners) like in the proposal of Pierrick. This is now fixed in
the trunk version (GWTEventService 1.3).
Could someone please confirm/check if the fix is completely solving the problem?
@JK: A client-side timeout handling could get implemented together with the
feature of Issue46 which is also planned for GWTEventService 1.3.
Original comment by sven.strohschein@googlemail.com
on 8 Aug 2012 at 10:32
Original issue reported on code.google.com by
Peter.Nu...@googlemail.com
on 6 Apr 2011 at 6:42