On Mon, 15 Sep 2003, Andreas Rieke wrote:

> as I am not familiar with cURL debugging, maybe a tcpdump can help.

I don't think so, as your description of the sequence was enough to hint about
that there's something fishy going on. We need to step through the code to see
what libcurl thinks.

> If you like, I can also send my application (it works on linux) which
> connects to our public server so you should be able to reproduce the
> problem. Which way do you prefer? Or can you provide a special debug
> libcurl.so.2 which writes debug information which are helpful to find the
> problem?

The best option for me, would be if you could produce an app that can
reproduce this problem with as little extra weight as possible, and then make
that available to the list or to me personally. We could then single-step
through libcurl to see why it doesn't properly detect and close the first
connection.

I personally am pretty swamped right now so it might take a while before I can
proceed on this. I don't expect my work load to reduce in the nearest upcoming
months either.

(I'm also on a backlog of other reported bugs.)

-- 
 Daniel Stenberg -- curl: been grokking URLs since 1998
-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf