>=daniel=>
>This means that the following program is legal in C++ and
>ANSI C, and has different meanings in the two languages.
>Am I correct in thinking that a design goal of Bjarne's was
>to avoid the possibility of such a program?
>Is that design goal sufficiently important to motivate x3j16
>to change the type of 'c' to int?
>[... program exercising sizeof('a') ...]

Yes and no.
Yes, being consistent with C would have been a strong enough reason
for Bjarne to have made the type of 'c' be int in the first place,
EXCEPT that there is a stronger reason to have 'c' be char in C++:
overloading.

If I write:

 cout << 'A';
 cout << 65;

I want overload resolution to choose operator<<(ostream&, char) for
the first call and operator<<(ostream&, int) for the second.  If 'A'
was type int then the first call would print the int value of A rather
than the character A.  That would be very ugly.

This is discussed by E&S (ARM) on p.9, 2.5.2, in the annotations.

=======================================
== Brian M. Kennedy <bmk@csc.ti.com> ==
== Computer Systems Laboratory ========
== Computer Science Center ============
== Texas Instruments ==================

In article <1991Jun18.210440.16325@csc.ti.com> bmk@m2.csc.ti.com (Brian M Kennedy) writes:

> . . . being consistent with C would have been a strong enough reason
> for Bjarne to have made the type of 'c' be int in the first place,
> EXCEPT that there is a stronger reason to have 'c' be char in C++:
> overloading.
>
> If I write:
>
>  cout << 'A';
>  cout << 65;
>
> I want overload resolution to choose operator<<(ostream&, char) for
> the first call and operator<<(ostream&, int) for the second.  If 'A'
> was type int then the first call would print the int value of A rather
> than the character A.  That would be very ugly.
>
> This is discussed by E&S (ARM) on p.9, 2.5.2, in the annotations.

The reason for 'a' to be a char is indeed for overload resolution to work,
as discussed in the ARM in 2.5.2, but the example given above is not a good
example.  In fact, in the implementations I'm familiar with (and I haven't
used 2.0, so I don't know if it's still true),

  cout << 'A'

prints the int value of A.  See "The C++ Programming Languange" by Stroustrup,
1987 edition.  On p. 9, 8.2.1, we see class ostream defined with
operator<<(char*) and operator<<(int), but no operator<<(char), and two pages
later, on page 229, we see
----------------------------------------------------------------------
   There are no character valued expressions in C++.  In particular, '\n' is
an integer (with the value 10 when the ASCII character set is used), so that

       cout << "x = " << x << '\n';

writes the number 10 and not the expected newline.
----------------------------------------------------------------------

The ARM does not address this issue, leaving ostreams, cout, and overloaded
<< operators as a library issue rather than a language issue.
What do 2.0 implementations do with this?  Is this an incompatibility that
will cause problems for C++ 1.2 programs passing "char"s to << when the
programs are compiled with C++ 2.0?
------------------------------------------------
Morris M. Keesan   mmk@d62iwa.mitre.org