endless loop in compute_inverse bootstrapping on alpha

[Brad Lucier]

> I'm appending the obvious extension of my previous patch to also change the 
> type of inverse_table[order].mult to size_t.  It just passed the critical 
> point in the bootstrap.  Next I will test Zack's patch, so whatever will be 
> applied will be confirmed to work as intended on alpha, also.

I applied your first patch and ran a bootstrap beyond a good ompare of the
stage2 and stage3 files.

I applied Zack's patch and got a complete bootstrap, and make check is
running now.

I'm inclined to let someone else check your last patch ;-).

Brad