Re: Fun with C

[Georg Bauhaus <rm.dash-bauhaus@futureapps.de>]

On 4/16/11 11:12 PM, Ludovic Brenta wrote:

> Out of curiosity, is this actual result a predictable consequence of the
> language definition, or is it undefined behavior?

C's definitions of what happens to arithmetic operands look
like an interesting exercise in making definitions (work).
The standard seems to land the resulting rules somewhere
between Ada types and the effect of Ada.Unchecked_Conversion.
It is not unchecked conversion, I think.

C's rules do not assume a specific bit representation of
arithmetic types, IINM. An interesting consequence  is
that the C99 Rationale lists specific circumstances in which
a result "must be dubbed _questionably signed_". But,
"(...) of course, _all of these ambiguities can be avoided
by a judicous use of casts_."
-- C99 Rationale,  �6.3.1.1

A difference between C and Ada here will be the difference
between *can* and *must*, as in "can be avoided using casts"
and "must be stated using conversions".  Consider programmers
shouting

  I can!

or

  I must!

If the first means "I can do it!" and latter means "It forces me
to do that...", this puts humans in two opposing camps.  Guess who
will be considered to have the more positive attitude. :-)

(Placement in this or that group will not be based on a consistent
set of facts, of course, but with such a set, we we'd have much
less to shout at each other...)