Mac Forum / Programming / CodeWarrior / March 2004

It seems to me that the two examples below
violate Section 13 of Stroustrup's Annotated Reference Manual.
Probably I am mistaken.
Can someone explain?
(I am using CodeWarrior Development Studio for Mac OS v9)

Example 1:
struct J { public:
 J()             { m= 0;};
 void setVal(float) { m=1;};
 void setVal(short) { m=2;};
 short  m;
};

{
  J j1,j2,j3;
  j1.setVal(1.0);      // compiler errors: ambiguous call...
  j2.setVal(1);        // ...to overloaded function
  j3.setVal((float)1.0);     // this call ok
}    

why does the compiler not treat '1' as integral,
and "1.0" as floating point?

"an exact match is better than any conversion" - Stroustrup, page 313.

Example 2:
struct K { public:
 K()                   { m= 0; };
 void setVal(const float &)  { m=1;};
 void setVal(short &)     { m=2; };
 short  m;
}

"It is .. possible to distinguish between const T&, volatile T&,
and plain T&"- Stroustrup, page 308.
Nevertheless, in the the following code all four calls to 'setVal'
assign '1' to 'm', indicating that the compiler is treating
bot "1"' and "(short)1"as floating point values!
(if "const float" and plain "float" are distinguishable,
surely "const float" and plain "short" are distinguishable?)

{
  K k1,k2,k3,k4;
  k1.setVal(1.0);
  k2.setVal(1);
  k3.setVal((float)1.0);
  k4.setVal((short)1);
}