to multi_runsingle() without it being really necessary or good

timeout'ed connections and possibly deal with them too

HAVE_WINDOWS_H, HAVE_WINSOCK_H, HAVE_WINSOCK2_H, neither
HAVE_WS2TCPIP_H when __CYGWIN__ is defined.

fast treatment of many simultaneous transfers

not always zero!

used anymore since multi->num_alive was introduced

Silence compiler warning 'unused parameter running_handles' in function multi_runsingle(). This is done here returning multi->num_alive in the running_handles parameter even when functions that call multi_runsingle() at this moment overwrite the returned value with the one that is valid when those functions curl_multi_perform() and multi_socket() have removed expired timers from the splay. Most probably, parameter 'running_handles' in function multi_runsingle() should be just removed.

*multi_socket*() can't return the proper number

if the header "Transfer-Encoding: chunked" was set by the application.
http://curl.haxx.se/bug/view.cgi?id=1531838

an unknown error number on glibc systems.
http://curl.haxx.se/bug/view.cgi?id=1532289

both now provide the number of running handles back to the calling function.

Winsock and Cygwin need address family specification before bind(), this should be harmless for others.

Reintroduce checking for HAVE_MSG_NOSIGNAL in configure script, so that we don't depend on header inclusion order for a valid check.

formposts work exactly the way you want it (and the way you'd assume it
works)

behaviour and a socket leak