PATCH COMMITTED: Don't break tests for enum in range

[Eric Botcazou]

> Because I think it will draw the correct conclusion that the value is
> out of bounds, rather than producing a bad range which leads to the
> conclusion that the value is in bounds.

I was talking about the non-inlined version of the function.

> Well, the choices of TYPE_PRECISION and TYPE_MAX_VALUE are clearly not
> accidents.  Whether they are correct or not is the type of question
> which might be addressed if we had a formal GIMPLE type system.  But
> even if we did, it might or might not help this case, since fold is
> being called by the front end.

So now objects can have values outside of their type?

> Sounds reasonable.  I certainly have no objection to that patch if it
> handles the new test case as expected.

OK.  The intent of my patch was to return the empty range, assuming that the 
arguments it is given are valid.  Forcing the high bound to NULL will make it 
more robust instead of disabling it.

-- 
Eric Botcazou