Re: parser bug for numeric constants

The parser gets the type for floating-point numeric constants wrong.  
The constant


is typed as T_FLOAT, whereas it should be T_DOUBLE (Ellis &  
Stroustrup section 2.5.3, "The type of a floating constant is  
_double_ unless explicitly specified by a suffix".

Yes, indeed, the parser gets it wrong. What's worse it seems to drop  
a suffix if there is one. I'll look into this. Thanks for pointing it  
out.   --Beata