#651464.

for a bigger buffer and this is indeed badness for us. Making this work
on both old and new glibc versions require an ugly loop that in its worse
form cause 45 bad loops when using the correct glibc and a non-resolving
host name... :-/

We want a better fix. Badly.

or not

multi interface works with file:// URLs fine (previously it crashed). This
won't make it work on Windows though...

regular transfer process. This required some new tweaks, like for example
we need to be able to tell the tranfer loop to not chunky-encode uploads
while we're transferring the rest of the request...

read callback, and thus won't put a lot of stress on the request sending
code (which currently does an ugly loop).

connectdata struct instead, and is no longer modified within the 'set' struct
as previously (which was a really BAAAD thing).

returning back to the select() loop. Consider this a test so far.

complete is now only made for 60 seconds and if no data was received during
those 60 seconds, we store a special error message (preparing to make this
a special error code) as this most likely means that the control connection
has died while we were transferring data.

proper CURLcode. The default timeout for reading one response is now also
possible to change while running.

chunked transfer work

The chunked transfer upload never stopped due to a silly add before we checked
for >0!

We now check if the chunked transfer-encoding header has been added "by force"
and if so, we enabled the chunky upload!

FOLLOWLOCATION and CURLOPT_USERPWD.

for freeing single dns cache entries

for storage and retrieval of private data in the curl handle.

ConnectBits where the other bools already are

AGGRESIVE_TEST is also defined, as an addition to MALLOCDEBUG. It doesn't
work for multi interface usage and should only be used with careful
consideration.

size of the uploaded file is unknown.

CURLOPT_CONNECTTIMEOUT was used and provided a fix.

debug wasn't enabled