MIPS SIZE_TYPE and PTRDIFF_TYPE revamp

[Eric Christopher]

> In retrospect, are you sure this patch fixes the problem?
> 
> AFAICT, the problem is still there w/ current sources as of this
> afternoon.  (to refresh memories: target=mipsisa32-elf.)
> 

I thought it did, but I'll take a look. I assume you did a fresh
configure and rebuild (or at least touched tm_p.h)?

> 
> BTW, is there any reason these things aren't done using a similar
> style as the alpha cpu specs are done?  (it looks like there's a _lot_
> of unnecessary duplication here, and getting rid of that would help
> both readability, maintainability, and properly correctness.  See the
> code in alpha/alpha.h starting w/ a comment w/ the words "Attempt to
> describe CPU" for an indication of what i mean.)
> 

Yes, unfortunately this field is dependent on a number of different
things (abi, isa, long). The ideal situation is to be able to base the
value on Pmode via the backend, however, until the integrated save-temps
work is completed we won't be able to use the backend to describe this.

It's conceivable that we can do this via the same mechanism - I may give
it a try.

-eric

-- 
I will not use abbrev.