Bug 1266 – Cannot forward reference the typeof of the base type of a pointer whose base type is defined with typeof

One of those hard to summarize legibly, yet relatively simple, bugs: void f(typeof(*y) z) {} typeof(x)* y; int x; Replace typeof(x) or typeof(*y) with int or move f after y and the code compiles. As it is, it emits "Error: error: forward reference of typeof(x)" without a line number for the error, in both DMD 1.014 and 1.015.

I was going to file this as a separate Bug, but it seems it's just a wonky error message for essentially the same thing: void f(typeof(*y) z) {} typeof(&x) y; int x; The above code errors out with the senseless "Error: can only * a pointer, not a 'typeof(&x)'", but at least it comes with a line number pointing at the typeof(*y), unlike the original code which used typeof(x)*.

Further testing: f isn't needed, "typeof(*y) z;" is enough. However, in the original ("typeof(x)* y;") case, using only "typeof(*y) z;" without f causes a compiler crash (no error message). Hence I'm stepping up the severity.

At least it gives a line number now.

This seems to be fixed on DMD2.028 (although possibly because of some other patch I've made <g> -- that would indicate that it's a duplicate). Still segfaults on D1.042.

I've fixed the seg fault in the upcoming 1.045 version, but the forward reference error remains.

This is no longer an ICE, just one of the many rejects-valid forward reference issue.

Here's a patch, but I am not 100% sure, whether the scope is correct for the type aswell. Index: expression.c =================================================================== --- expression.c (revision 421) +++ expression.c (working copy) @@ -7432,6 +7432,7 @@ error("can only * a pointer, not a '%s'", e1->type->toChars()); return new ErrorExp(); } + type = type->semantic(e1->loc, sc); // correct scope? rvalue(); } return this;

Compiles without error on 2.052 and 1.067

Fix confirmed: this failed in 2.050 and 1.065 but is fixed in 2.051 and 1.066