AC_CHECK_TOOL is prolly better to use when checking for ar

slightly odd fix to prevent -Wunreachable-code to warn

alert the user if 'sed' or 'ar' couldn't be found, as it might very well
render a build impossible

made the progress meter display not overflow even if _very_ large files
are transfered. The maximum size we support now is 8 exabytes, which equals
to 8192 petabytes...

if the values allow it, avoid floting point math for the current speed

additional typecasts in an attempt to avoid compiler warnings when switching
from 64 bit types to 32 bit ones

removed bad free()

do the alarm time-left math using unsigned longs since that is what alarm()
returns and uses as input and converting to signed generates warnings and
actually risks loss of accuracy

fix_hostname() now (void)s the conn argument to prevent warnings on non-idn
enabled builds

ERR_error_string() returns an unsigned long so we should use one of those
for the return code

gcc 3.4 now uses the -Wunreachable-code option, I believe we can make older
ones use this too...

check the size of size_t for lib/mprintf.c

prevent warnings when using the gcc option -Wunreachable-code

make the memlimit final NULL return get written to stderr as wella

mention the LICENSE-MIXING document

minor update edits

added a third URL to the torture testing, this one also hangs at some point
for a reason I don't know

multipart formposts should be more streamy

bail out when an add_buffer() function returns failure

improved the cleaning up of memory when we fail to resolve names due to
out of memory (thanks to 'runtests.pl -t')

check malloc() return code

removed more leftovers from the formparse function

removed curl_formparse() from the library

Gisle made item 38, now there's only one low-prio task left...

General HTTP authentication cleanup and fixes