When adding entries to the DNS cache with CURLOPT_RESOLVE, they are
marked 'inuse' forever to prevent them from ever being removed in normal
operations. Still, the code that pruned out-of-date DNS entries didn't
care for the 'inuse' struct field and pruned it anyway!

Reported-by: Romulo A. Ceccon
Bug: http://curl.haxx.se/bug/view.cgi?id=1303

Removed unnecessary SMTP_STOP state changes on failure.

Removed hard return on failure in smtp_state_data_resp().

Added new API for returning a SSL backend type and pointer, in order to
allow access to the TLS internals, that may then be used to obtain X509
certificate information for example.

Commands such as NOOP, RSET and HELP would be sent with a space at the
end of the command, for example: "NOOP ".

Otherwise a NOOP operation would be performed which a) only returns a
single line response and not a multiline response where -I needs to be
used, and b) provides an inconsistent user experience compared to that
of the POP3 and IMAP protocols.

The buffer allocated by smtp_parse_custom_request() was not freed.

singleipconnect() could return the file descriptor of an open socket
even though the function returned a CURLE_COULDNT_CONNECT error code
from commit ed1662c3 and 02fbc26d.

This could cause tests 19, 704 and 1233 to fail on FreeBSD, AIX and
Solaris.

Incorrectly processed multiline server greeting responses as "wanted"
continue responses in smtp_endofresp(), from commit f16c0de4,
which in turn broke the SMTP server detection in the test suite,
because the EHLO command would not be sent.

warning: unused variable 'smtpc'

Added a loop to smtp_statemach_act() in which Curl_pp_readresp() is
called until the cache is drained. Without this multiple responses
received in a single packet could result in a hang or delay.

Similar to the processing of untagged CAPABILITY responses in IMAP moved
the processing of multiline EHLO responses to smtp_state_ehlo_resp() and
introduced an internal response code of one to differentiate a multiline
continuation from the end of command. This also allows for the separate
processing of multiline responses from commands such as VRFY and EXPN.

singleipconnect() did not return the open socket descriptor on some
errors, thereby sometimes causing a socket leak. This patch ensures
the socket is always returned.

Even though this is only a formality (since not many people build on
Mavericks while targeting Leopard), since we still support Leopard
at the earliest, we might as well be pedantic.

It turns out that some of the constants necessary to make this feature
work are missing from Snow Leopard's Security framework even though
they are defined in the headers.

Bug: http://curl.haxx.se/mail/lib-2013-11/0076.html
Reported by: myriachan

Rather than set the authentication options as part of the login details
specified in the URL, or via the older CURLOPT_USERPWD option, added a
new libcurl option to allow the login options to be set separately.

This patch fixes and issue introduced in commit 7d7df831, if the
tunnel state was TUNNEL_CONNECT, waitconnect_getsock() would return a
bitmask indicating a readable socket but never stored the socket in the
return array.

The temporary sockets used for Happy Eyeballs were not closed properly,
if curl exited prematurely, which this patch fixes.

The reason for this loop's existence was removed in commit
02fbc26d.

Our own printf() replacement clearly can't properly handle %.*s with a
string that isn't zero terminated. Instead of fixing the printf code or
even figuring out what the proper posix behavior is, I reverted this
piece of the code back to the previous version where it does malloc +
memcpy instead.

Regression added in e839446c, released in curl 7.32.0.

Reported-by: Felix Yan
Bug: http://curl.haxx.se/bug/view.cgi?id=1295

This patch adds a 200ms delay between the first and second address
family socket connection attempts.

It also iterates over IP addresses in the order returned by the
system, meaning most dual-stack systems will try IPv6 first.

Additionally, it refactors the connect code, removing most code that
handled synchronous connects. Since all sockets are now non-blocking,
the logic can be made simpler.

nss.c:702: warning: pointer targets in passing argument 3 of
'Curl_extract_certinfo' differ in signedness

nss.c:702: warning: pointer targets in passing argument 4 of
'Curl_extract_certinfo' differ in signedness

Made sure the cast was correctly "unsigned char *" to "char *" and not
"unsigned char *" to "unsigned char *".

nss.c:700: warning: pointer targets in passing argument 3 of
'Curl_extract_certinfo' differ in signedness

nss.c:700: warning: pointer targets in passing argument 4 of
'Curl_extract_certinfo' differ in signedness

No need for a rhs condition on a bitwise compare.

warning: implicit declaration of function 'Curl_extract_certinfo'

warning: declaration of 'chsize' shadows a global declaration

curl_sasl.c:294: warning: dereferencing type-punned pointer will break
strict-aliasing rules

getpart.c:201: warning: dereferencing type-punned pointer will break
strict-aliasing rules

Introduced in commit 7d7df831 curl would loop displaying "Whut?"
if it was trying to connect to an address and port that didn't have
anything listening on it.

Renamed copy_header_value() to Curl_copy_header_value() as this
function is now non static.

Simplified proxy flag in Curl_http_input_auth() when calling
sub-functions.

Removed unnecessary white space removal when using negotiate as it had
been missed in commit cdccb422.