[klibc] Build conundrum solved - sort of

[Bryan O'Sullivan]

I've figured out what was going on that was making it impossible for me
to build klibc.  There may have been some version drift between the
klibc README and how things work in 2.5, but in order for
..../linux/include/asm to be created under an otherwise empty 2.5.68
tree (i.e. one that hasn't seen a build), one must run one of the make
configs, followed by a "make prepare".

This gets things going, but there's still an issue with the klibc build
in 0.80; the socketcalls subdirectory doesn't contain anything, so the
call to "ar cq" fails.  The issue seems to be that the big "is this a
prototype?" regexp on line 8 of socketcalls.pl is broken, so the script
doesn't generate any socketcall stubs.

This appears to be a more robust formulation of the same regexp:

   $line =~ /^\s*(.*)\s+([_a-zA-Z]\w+)\s*\((.*)\)\s*$/

With this in place, the 0.80 snapshot now builds fine.

By the way, it may be worth also noting in the klibc README, for the
feeble-minded such as myself, that if you're building against a pristine
kernel BK snapshot, it's necessary to do a "bk -r get -q" in the kernel
tree (which checks out all files) *before* the "make *config" and "make
prepare" steps, otherwise klibc won't be able to find most kernel
headers.

	<b