ares_free_errmem is gone

unused and now removed

Removed getpass from here. The fact this still was present here indicates
that this file is out of date!

curl_off_t is 8 bytes big on windows

switch the arguments according to the c source

oops, the decode() function got its arguments reversed in my cleanup
operation!

correct the input data structs

use size_t to keep strlen() results

simplied how create_hostcache_id() is used, and also its function somewhat
cleared up some ssize_t/size_t mixups

adjusted to the modified base64 protos

More size_t cleanups in the base64 functions.

The base64 encode function now takes a size_t for size, not an int as
previously.

more int vs long/size_t fixes after icc compiler warnings

more int/long fixes after icc "remarks"

next_id is an unsigned short, typecast the assign to prevent picky compilers
to warn

typecase the bitfiddling results since we get a long and we store an int,
they may not be of the size size

make ares_expand_name() take a long * instead of an int *, since we do
pointer arithmetic (ptr1 - ptr2) and to do that properly on 64bit we need
long

strlen() returns a size_t, which might be larger than int on some platforms

argh, use single-quotes instead of double ones to make $Revision appear

use the proper source header, and set the version string to this file's
revision number

cd back to the root path before removing the build dir, since some systems
refuse to remove the dir otherwise!

Dominick Meglio's fix for supporting multiple names in the Nameserver key
on Windows.

this code uses no long long, so we can have warnings about them

the missing part of Gisle Vanem's connect-timeout fix for win32

-Wno-format-nonliteral does not exist in my gcc 2.96, only use that with
newer versions