Hello.

I found another bug in pipelining. Steps to reproduce:

1. Add the following postcmd to reply section of test530 data:

<postcmd>
wait 3
</postcmd>

  This makes test530 fail due to bad handling of pipeline requests in
sws (I plan to fix it in another patch).

2. ./runtests.pl -k -v 530

  It timeouts and segfaults while doing cleanup.

The problem is that when curl_multi_remove_handle() is called for
pipelined requests in lib530.c:167, easy handle does not "own"
connection. Check at multi.c:612 fails and Curl_done() is not called.

Easy handle is left in pipeline, and we call cleanup for it at
lib530.c:168. Later when signalPipeClose() gets to this handle we get
"signalPipeClose() found BAAD easy handle" message and segfault.

I am not sure this is correct solution. Counting on you here, Daniel.

Regards,
  Dmitry