From: "Luke-Jr" <luke@dashjr.org>
To: bitcoin-development@lists.sourceforge.net
Date: Fri, 9 Sep 2011 02:57:03 -0400
User-Agent: KMail/1.13.7 (Linux/2.6.39-gentoo; KDE/4.6.5; x86_64; ; )
MIME-Version: 1.0
Content-Type: text/plain;
  charset="us-ascii"
Content-Transfer-Encoding: 7bit
Message-Id: <201109090257.06644.luke@dashjr.org>
Subject: [Bitcoin-development] 0.3.23+patches bug: JSON-RPC leaves sockets
	around when not connected
Precedence: list

On Eligius, I have two bitcoinds running on the same system:
- a hub node, which is dedicated to relaying network activity between the
  hundreds of nodes Eligius peers with
- a work node, which is dedicated to managing mining, and only ever connects
  to the hub node

Lately, the hub node has been dying (yet to determine cause; I haven't changed 
anything recently), leaving the work node with no connections. The work node, 
of course, starts responding to getworks with "not connected" messages. 
Strangely enough, the sockets used for these are never cleaned up properly, 
and eventually overflow the system's socket limits.

I am using a very heavily patched 0.3.23, so this bug could be related to one 
of those patches (perhaps the threaded JSON-RPC?), but I felt like I should 
mention it somewhere else just in case before 0.4 final got released. I have 
not tried to reproduce it on an unpatched bitcoind yet, and am extremely busy 
lately so I might not get a chance before you guys finish 0.4.