transport crashes: can't start new thread

[GoogleCodeExporter]

What steps will reproduce the problem?
1. After a while of using the transport. It's not clear why.
2.
3.

What is the expected output? What do you see instead?
Message send to my client from the transport:

  "Undefined condition.
  The error condition is not one of those defined by the other conditions
in this list.
  ------
  ICQ connection lost! Reason: [Failure instance: Traceback (failure with
no frames): <class 'thread.error'>: can't start new thread
  ]"

What version of the product are you using? On what operating system?
pyicqt 0.8.1.3, Debian 5.0, Python 2.5.2, Twisted 8.1.0, ejabberd 2.0.3

Please provide any additional information below.
After a while I had used the transport, it crashes with the error-message
I've written above. After the error occurs, the transport is still running,
but he will not accept any new connections. Only restarting the transport
is not enough. To get the transport working, I had to reboot the whole OS.

Original issue reported on code.google.com by gm...@pf0hl.de on 17 Jun 2009 at 10:36

[GoogleCodeExporter]

Looks as OS-related problem. May be you have no enough virtual memory or other
resources? PyICQt use not many threads, ~30 in my testing installation with 
some users

Original comment by r000ns...@gmail.com on 7 Aug 2009 at 9:09

[GoogleCodeExporter]

This happens to me too since Gentoo upgrade to python-2.6, rolling back Python 
to 2.5
lowers memory&thread usage from 1300Mb+ to 900Mb (20 connected users, still too 
much
- but it runs without warnings and errors).

Original comment by Nikolaus...@gmail.com on 16 Aug 2009 at 4:38

[GoogleCodeExporter]

Sorry for comment spam, just noticed after a while (minutes, not seconds) it 
gets
down to about 100Mb with python 2.5, memory&thread-usage perhaps 
startup-related?

Original comment by Nikolaus...@gmail.com on 16 Aug 2009 at 4:39

[GoogleCodeExporter]

With 2 users it eats 340 Mb of virtual memory and 340 Mb in peak. :) You got 
maximum
when restarted transport I think and all users were connecting to it 
simultaneously.
Or 900 Mb in resident memory?

Original comment by r000ns...@gmail.com on 16 Aug 2009 at 6:27

[GoogleCodeExporter]

Now it's here early morning, much less users connected - tried pyicq-t with
python-2.6 on gentoo again, result with 9 connected users:
Memory usage during start 450Mb, 2 minutes later 116Mb, 9 threads.
With python-2.5 it uses a bit less memory, about 100Mb.
Yes, you're right, waiting some time reduces the memory usage to normal levels, 
and
if there are not too much users connecting at the same time (during startup) 
starting
the transport works without problems.
Small workaround: I removed/commented out <enableAutoInvite/> from my config 
file,
now it starts perfectly :) (but all users have to reconnect themselfes)

Original comment by Nikolaus...@gmail.com on 17 Aug 2009 at 6:32

[GoogleCodeExporter]

Try 0.8.1.5 version

Original comment by r000ns...@gmail.com on 24 Aug 2009 at 5:04

[GoogleCodeExporter]

the new version feels better :) Tested with 15 users connected and
<enableAutoInvite/> activated, i get this results:
0.8.1.4    highest memory usage during startup: 880 Mb, reduces to 190 Mb
0.8.1.5    highest memory usage during startup: 781 Mb, reduces to 125 Mb

Original comment by Nikolaus...@gmail.com on 25 Aug 2009 at 6:47