• Matthew Wilson (20/20) Sep 22 2002 Not sure whether this should only be a warning (haven't mustered the eff...
• Walter (7/26) Sep 23 2002 effort
• Matthew Wilson (2/4) Sep 23 2002 In which case I prefer the DMC version. Helps me that bit more to valida...

"Matthew Wilson" <dmd synesis.com.au> writes:

Not sure whether this should only be a warning (haven't mustered the effort
to trawl the standard this early in the morning)

class X
{
public:
    operator const char const * () const
    {
        return 0;
    }
};

gives

 Error: ...: illegal combination of types

I expected it to simply tell me that the first (or second) const was
superfluous.

I don't object to this one in its current form, but am concerned as to
whether it means that the compiler would balk in a template class defintion
where that template defines a member typedef an a constant type with which
it is parameterised, and the parameterising type is itself const. Borland
does have problems in this area.

Matthew

"Walter" <walter digitalmars.com> writes:

"Matthew Wilson" <dmd synesis.com.au> wrote in message
news:amlfgi$p8$1 digitaldaemon.com...
 Not sure whether this should only be a warning (haven't mustered the

effort
 to trawl the standard this early in the morning)

 class X
 {
 public:
     operator const char const * () const
     {
         return 0;
     }
 };

 gives

  Error: ...: illegal combination of types

 I expected it to simply tell me that the first (or second) const was
 superfluous.

I believe the standard says it's supposed to be an error, just like short
short is an error, and short int short is an error.


 I don't object to this one in its current form, but am concerned as to
 whether it means that the compiler would balk in a template class

defintion
 where that template defines a member typedef an a constant type with which
 it is parameterised, and the parameterising type is itself const. Borland
 does have problems in this area.

This should work fine.

"Matthew Wilson" <dmd synesis.com.au> writes:

 I believe the standard says it's supposed to be an error, just like short
 short is an error, and short int short is an error.

In which case I prefer the DMC version. Helps me that bit more to validate
the code.