Simplify check for NEED_MALLOC_H, and make more explicit that NEED_MALLOC_H shall be defined if <malloc.h> header file must be included even when including <stdlib.h>.

Define NEED_MALLOC_H if including <stdlib.h> is not enough for proper compilation and <malloc.h> must also be included.

Finally get rid of CURL_CHECK_HEADERS_ONCE since it adds very little value and has portability issues.

Change some shell if...then...fi tests into case...esac tests which demand less resources.

Using backslashes and slashes in the strings of the sed 'y' command shall be avoided since its interpretation is not the same across platforms.

Now we use the sed 's' command with a bracket expression.

Prevent definition of HAVE_WINxxx_H symbols and avoid inclusion of Windows headers when compiled with Cygwin in POSIX emulation mode.

HAVE_GETNAMEINFO, GETNAMEINFO_QUAL_ARG1, GETNAMEINFO_TYPE_ARG1, GETNAMEINFO_TYPE_ARG2, GETNAMEINFO_TYPE_ARG46, GETNAMEINFO_TYPE_ARG7

HAVE_RECV, RECV_TYPE_ARG1, RECV_TYPE_ARG2, RECV_TYPE_ARG3, RECV_TYPE_ARG4, RECV_TYPE_RETV

HAVE_SEND, SEND_TYPE_ARG1, SEND_TYPE_ARG2, SEND_TYPE_ARG3, SEND_TYPE_ARG4, SEND_TYPE_RETV

ares_dns.h, which break c-ares on my Sparc64. Bit-wise operations in C
operate on logical values. And in any event the octets are already in
big-endian (aka network) byte order so they're being reversed (thus the
source of the breakage).

UDP even when a poll(2) or select(2) suggest otherwise.

libtool cross-compiled on linux with mingw32

things such as C++ compiler actually is a bad thing and since we don't need
that detection I added a work-around, much inspired by a previous patch by
Paolo Bonzini. This also shortens the configure script quite a lot.

moved the curl_off_t check to within the --enable-debug block where it belongs since it is a somewhat ugly hack