proposed patch for constant signed/unsigned comparison warning

[Joe Buck]

I think that Paul's analysis should settle the issue:

> Linus Torvalds originally brought up this problem in May 1996 with
> respect to the Linux kernel.  I recall that rms showed some sympathy
> for his complaint at the time.  I also recall that kenner wrote that
> he was thinking of making GCC smarter so that it didn't generate these
> warnings when they were bogus.  That would be nice, but it hasn't been
> done yet and it's not trivial to do.  Until GCC is made smarter, -Wall
> should not generate these warnings.

The warnings shouldn't be implied by -Wall until there is smarter
range analysis that can distinguish bogus warnings from true ones.