Re: just a little nitpik in bfd

[Nick Clifton <nickc at cambridge dot redhat dot com>]

Hi Stuart,

> In the core of bfd (binutils 2.11.92), in file hash.c in function 
> bfd_hash_lookup, I see:
>  
> >  hash = 0;
> >  len = 0;
> >  s = (const unsigned char *) string;
> >  while ((c = *s++) != '\0')
> >    {
> >      hash += c + (c << 17);
> >      hash ^= hash >> 2;
> >      ++len;
> >    }
> >  hash += len + (len << 17);
> >  hash ^= hash >> 2;
>   
> Surely, we can eliminate ++len from the loop, because it is an
> induction variable (and the compiler might do it for us, but we can
> and should code carefully first, then hope the compiler can do
> better): 

Certainly - that would be a reasonable change - I will apply the patch
shortly.

> One more wrinkle: why do we need to hash len into the hash total?

It does no harm, and it might help in situations where a long string
with low numbered ascii characters has a similar hash value to a
shorter string with higher numbered ascii characters.

Cheers
        Nick