> > GC> x=new unit[]={...};

 > RM> The above line won't work because you're trying to initialize a
 > RM> pointer, not a structure. The line

 > DN> ??

 > It's the 2nd part of the statement I was pointing out, the "={...}"
 > bit. He's trying to initialize a pointer with structure data,
 > without dereferencing the pointer.


Ok.


 > RM> The fact that you have the structure inside a class only
 > RM> complicates things; the real problem is your attempting to
 > RM> initialize a pointer with structure contents.

 > DN> C++ can handle this situation just fine.

 > yes, but if you stuff up somewhere, the compiler may sometimes
 > give an error msg about a class violation or similar, rather than a
 > more accurate msg that you'd get outside the class.

Indeed. Error parsing is difficult at best, and tricky as hell to get
right. :-) C++ is not an easy language to parse.


 > Probably the answer I should have given Griffon is something like

 >  // In the class definition, set up y as a data member
 >  unit *y;

 >  // Somewhere else, point y at a new unit and initialize it
 >  y = new unit ['c',1];
 >  // and don't forget to delete(y) before the class disappears...

This won't work either.

What you'd need to do is to define a constructor for class/struct unit
which takes the correct arguments.


  David

---