On Fri, 30 Sep 2011, Vladimir Grishchenko wrote:

> Ok, I think I have a reproducible case in hand, which of course may be a
> user error. I'm going to try and attach the example and also copy it here.
> Look at the end of debug output, I don't know if reflects what is actually
> happening but it appears that an early cancellation leaves a connection
> alive despite the fact HTTP headers were already sent out. I see it
> happening with 7.22.0. Running on Win32.

I'm running this on Linux using my current git repo, but I fail to repeat it.
:-(

I always see the third request use a fresh connection after it was aborted by
a callback in the second request...

Any chance you can just set a break-point in lib/progress.c:356 and
single-step forwards a bit after your callback has returned non-zero and see
what makes this problem?

Also, please try a recent daily snapshot to see if we already have fixed this
problem. We just might've.

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