In article <3otggj$b4n@nntp5.u.washington.edu>
ghogenso@u.washington.edu (Gordon Hogenson) writes:

|> This question relates to overload resolution of functions
|> that take "long int" as an argument.

|> The problem is that the conversion "int --> long int" is
|> classified by the draft as a "conversion" rather than a
|> "promotion" (if I understand 4.5, conv.prom and 4.7,
|> conv.integral) correctly.  The result of this is that

|> int --> long int

|> is "just as good" a conversion as

|> int --> double

|> which is problematic in many common cases.

Correct.  But the rules are complicated enough as it is.

A good general design rule: if you have an overload for any of the
integral types, overload on int, too.

  Consider:

|> int pow(long int, long int);
|> double pow(double, long int);
|> double pow(double, double);

I would add:

 inline int
 pow( int i , int j )
 {
     return pow( (long)i , (long)j ) ;
 }

Simple.  No run-time overhead.  And it ensures that things like
`pow( 2 , 10 )' get the function you want.
--
James Kanze         Tel.: (+33) 88 14 49 00        email: kanze@gabi-soft.fr
GABI Software, Sarl., 8 rue des Francs-Bourgeois, F-67000 Strasbourg, France
Conseils en informatique industrielle --
                              -- Beratung in industrieller Datenverarbeitung

This question relates to overload resolution of functions
that take "long int" as an argument.

The problem is that the conversion "int --> long int" is
classified by the draft as a "conversion" rather than a
"promotion" (if I understand 4.5, conv.prom and 4.7,
conv.integral) correctly.  The result of this is that

int --> long int

is "just as good" a conversion as

int --> double

which is problematic in many common cases.  Consider:

int pow(long int, long int);
double pow(double, long int);
double pow(double, double);

the call:

int i;
double x;
pow(x, i);

is ambiguous.  If int-->long int were counted as an integral
promotion, then the call would resolve, which I believe is
the sensible result here.

Any comments?  Is this something that was thought about or is it
an oversight?


--
--
Gordon J. Hogenson <ghogenso@u.washington.edu>         Tel. (206) 685-2951