Corrupted keep-alive requests to server after adding attachment

[Javer]

When composing message, Roundcube periodically send keep-alive request to the server, which looks like below:

http://domain.tld/?_task=mail&_action=keep-alive&_t=1211398912731&_remote=1

But after uploading attachment new additional keep-alive thread are initialized, which use url like this:

http://domain.tld/&_action=keep-alive&_t=1211398974817&_remote=1

This generate HTTP 404 errors. After uploading another file, new one keep-alive thread with incorrect url are added and so on...

This occured because page downloaded by POST-request (during uploading file) contains initializing of new instance of rcube_webmail class, which create new keep-alive timeout, but use for this corrupted url, because there is no rcmail.set_env() call and script does not know which url to use.

var rcmail = new rcube_webmail();

I think, it is not necessary to create new instance of rcube_webmail in page-response to uploading file, because all changes are made through parent.rcmail.

[thomasb]

The upload form is posted into an iframe which is a separate window instance. I cannot imagine that new rcube_webmail() will override the object in the main window. Which browser did you test with?

However, in r1434 I just removed some unnecessary stuff from the response which should also fix this problem.

[Javer]

Problem was appeared at least in IE6 and 7 (can be detected using HttpWatch? plugin) and Firefox 2.x (using Firebug).
Revision 1434 solve this problem.