On Mon, 18 Feb 2013, Tom K. wrote:

> When the endpoint is reliable, this all works fine. When it is not
> reliable, it will lock up the system. Not immediately, though usually
> within 10 minutes or so.

200 is not a very large amount though. I've personally done 10K connections
with libcurl and I know others have done beyond 50K in a single machine.

I don't know of any such very big users on Windows though.

> I don't know what this really means in terms of what could be wrong. I'm not
> well versed in debugging issues like this, so any recommendations on
> troubleshooting this would be well received.

First, cut down the number of connections to the minimum that still repeats
the problem. Then, try to identify patterns on how to repeat the problem.

When everything "locks up" as you say, does the machine really stop working or
is there network traffic still going on?

To me, this sounds like a serious problem somewhere to a level that libcurl
shouldn't even be capable of doing. Like a network driver problem or so.

I'm totally clueless when it comes to how you can debug what Windows does when
it goes to this point. If you can repeat the problem on Linux however...

-- 
  / daniel.haxx.se
-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette:  http://curl.haxx.se/mail/etiquette.html