On Wed, 18 Oct 2006, David Laight wrote:
> This shows up another fubar in the ansi C 'value preserving' integral
> promotions (as opposed to the K&R 'signed preserving').

I don't think I've ever encountered a case where value preserving
is more useful than signed preserving.  In most cases, either the
difference is not noticeable, or signed preserving is more useful.  I
have no idea what the committee was thinking, breaking old code and
replacing useful behaviour with confusing behaviour.

And I can't find a GCC warning for "signed preserving is not what you
want here".

--apb (Alan Barrett)