linuxdcpp-team team mailing list archive

[Pirre <837544@xxxxxxxxxxxxxxxxxx>]

the problem is not that the socket is not closed on a disconnect or the
entity is removed, its the handle created by the application to manage
the socket that is not freed after the socket is closed.

and yes this will kill a hub, and even the complete OS as each OS is
limmited in max handles (NT based OS's geting into probs around 50.000
no idea on linux but guess thats higher lol).

a example 1 of the hubs i run reaches in 3 day's 11.000 handles just
caused by the fact that a lot of old non-adcs clients try to connect
over and overagain, if you get a flood of netbots who each connect @ a
400/min connection rate (and no a flood script can not prevent opening
the socket) any adch hub running the latest rev is a sitting duck :)

if i remember correct you changed the code so it called 2 methodes for
closing , resulting in a proper close of both handle and socket.

-- 
You received this bug notification because you are a member of
Dcplusplus-team, which is subscribed to ADCH++.
https://bugs.launchpad.net/bugs/837544

Title:
  Handles trapped on closing socket

Status in ADCH++:
  New

Bug description:
  Rev 519 tryed on both XP and Win2k8 server

  When a user logs out not all handles are freed (on logon it creates 3
  handles , also a bit strange) and when the socket is closed only 2 of
  them are freed resulting in a core that is notgoing to run long.

  This bug existed before untill poy changed the way the socket was
  closed.

To manage notifications about this bug go to:
https://bugs.launchpad.net/adchpp/+bug/837544/+subscriptions