On Wed, 7 Jun 2006, Patrick Heeb wrote:

> The strange thing is: After the timeout, the application retries the upload
> using the same easy_handle (* Re-using existing connection! (#0) with host
> legolas) and the server immediately connects to the client and the upload is
> done successfully.

Yes, but it makes another connect attempt with a completely different port
number etc so they're not identical. Can you see any other patterns beside the
first-fail second-succeeds?

To me it simply looks like the server doesn't connect back fine in the first
attempt but does so in the second. Can you ethereal the connect attempt (or
similar) to see what really happens on TCP level?

> Wed Jun 7 15:12:24 2006 [pid 15627] [bast] FTP command: Client
> "192.168.2.224", "PORT 192,168,2,224,196,163"
> Wed Jun 7 15:12:24 2006 [pid 15627] [bast] FTP response: Client
> "192.168.2.224", "200 PORT command successful. Consider using PASV."
> Wed Jun 7 15:12:54 2006 [pid 15627] [bast] FTP command: Client
> "192.168.2.224", "CWD /home/bast"

Any chance you can get to see what the server thinks about the connect attempt
it did back to the client due to the PORT command?

-- 
  Commercial curl and libcurl Technical Support: http://haxx.se/curl.html