Why will the following code compile when DG is an alias, but not when it is a unique type?  Can delegates not be typedef'd?  The compiler bawks at the assignment of the delegate, not the typedef--or even the member function returning DG.

Example code:
-----------------------------
version(wontcompile) {
typedef void delegate() DG;
} else {
alias void delegate() DG;
}

class A {
void foo() { printf("hi joe"); }
DG pfoo() { return &this.foo; } //this is ok
}

int main(char argc[][])
{
A a = new A;
//these won't compile with a strict type
DG dg = &a.foo;
dg();
DG dg = a.pfoo();
dg();
return 0;
}

On Fri, 23 Aug 2002 02:19:13 +0000 (UTC) Joe Battelle <Joe_member@pathlink.com> wrote:

> Why will the following code compile when DG is an alias, but not when it is a unique type?  Can delegates not be typedef'd?  The compiler bawks at the assignment of the delegate, not the typedef--or even the member function returning DG.

Of course. Typedef creates a new type, based on the one you specify. It will
be implicitly
convertable to that type, but you must use a cast to convert base type to
typedef:

	typedef int number;

	int n;
	number m;

	n = m;	// okay, number is "derived" from int
	m = n;	// error, should use cast()

Note, the compiler says "cannot implicitly convert int to number", but don't
let the buggy
erro message fool you - it actually complains about "m = n" (BTW seems like a
bug!).
This applies to all types, delegates included.

Yeah, I got that about typedefs.  Sorry, I reported the bug slightly wrong, the
message is: if.d(23): function expected before (), not 'DG'.

So it gets past the assignment.  Indeed no cast seems to be needed on either the return &this.foo or the assignment of dg=&a.foo.  Maybe the compiler is lax on type checking these. Certainly no cast needs to be made for dg=a.pfoo() for the return type is DG.

It looks like the compiler says what is has to the left of the function call operator is not a function, and indeed it's not.  But it should be interpreted as one.