explicit array return syntax used in forward declaration generates broken assembly

[YashasSamaga]

Issue description:

Using the forward [n]funcname(args); syntax causes the compiler to generate the wrong code.

Minimal complete verifiable example (MCVE):

forward [4]f();
f() {
	new x[4];
	return x;
}
main () {
	new arr[4];
	arr = f();
}

generates

proc ; f
	stack fffffff0
	zero.pri
	addr.alt fffffff0
	fill 10
	addr.pri fffffff0
	const.alt 0
	movs 10
	stack 10
	retn

proc	; main
	stack fffffff0
	zero.pri
	addr.alt fffffff0
	fill 10

	addr.pri fffffff0
	push.pri
	heap 10
	push.alt
	push.c 0
	call .f
	pop.pri
	pop.alt
	movs 10
	heap fffffff0

	stack 10
	zero.pri
	retn

The array is not returned correctly. It instead corrupts the memory by writing to 0.

It should have instead been load.s.alt 12.

Workspace Information:

• Compiler version:
• Command line arguments provided (or sampctl version):
• Operating System:

[YashasSamaga]

The suspected piece of code involved in this bug is:

/* attach the array to the function symbol */
  if (numdim>0) {
    assert(sym!=NULL);
    sub=addvariable(symbolname,0,iARRAY,sGLOBAL,tag,dim,numdim,idxtag,0);
    sub->parent=sym;
    if (sym)
      sym->child=sub;
  } /* if */

Since the code that is generated when the compiler automatically deduces the returning array size works, I guess if the sub was set as a iREFARRAY with the offset as (argcount + 3) * sizeof(cell), it should work.

[YashasSamaga]

Should this even be fixed? There is no mention of this in the implementer/language guide and it does not provide any useful advantage (as far as I can see).

[Y-Less]

The only thing even vaguely related to this is explicit mention that public functions can't return arrays. forward can be used on other functions as well, but not native, which is the only one that needs the explicit array size. So we need to figure out the exact semantics of when this would apply.