better for FTP (any protocol with protocol-specific connect actions)

CURLOPT_DEBUGFUNCTION and CURLOPT_DEBUGDATA.

connect() so that it tells to which host and what port it is about to
attempt the connection

always stop reading after the headers no matter what 'close' is.

check proxy connections for "deadness" before they were re-used

extra second after having downloaded headers-only

Gerhard Herre reported this in bug report #536238

to make sure it remains functional.

array
Also, uploading data with --crlf cannot check that the proper size was
uploaded and thus should not warn if the sizes differ. This can be changed
in the future by having the expected size in the connectdata struct and then
increase the expected size for each byte that is added in the --crlf
replacement process.

and it cause cause a crash.

Albert Choy found and fixed it.

an error, even if we were decoding a chunked-encoded transfer

Now first check if IPv6 is supported, then use PF_UNSPEC. If not, use PF_INET.
It'll solve both the "slow name lookup" problem on IPv4 and still work fine on
IPv6 hosts.

Bug report #530204 has more details:
http://sourceforge.net/tracker/?func=detail&atid=100976&aid=530204&group_id=976

functionality and this change reverts this.

However, with this revert we bring back problems on (some/all?) non-IPv6
enabled Linux machines that have getaddrinfo().

we should read the POST-data from the read callback

(but didn't crash or anything)

do if the write would've blocked

part.