>> +  if (TYPE_NAME (a)&&  TYPE_NAME (b)
>> +&&  !strcmp (TYPE_NAME (a), TYPE_NAME (b)))
>> +    return 1;
>
> A minor stylistic nit: can we please use `strcmp (...) == 0' instead
> of `!strcmp (...)'?  It will produce the same code, but is slightly
> more readable.
>

Oh yeah, I noticed that too but forgot to change it.

This patch is revised for the above and contains a fix similar to the 
'do_is_ancestor' one pointed out by Tromey in patch 1 of this series. 
Some tests need to be update as a result.