Bug 284 – Wrong type of template value parameter

struct A(int x) { } template B(uint x) { static if (x == 3) static assert(is(A!(3) == A!(x))); } static this() { B!(3); } --- file.d(5): static assert (is(A == A )) is false --- Two instances of A!(3) is created. One with an int value type and the other one with a uint(!) value type. The mangled forms are *__T1AVi3Z1A and *__T1AVk3Z1A respectively.

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 [email protected] schrieb am 2006-08-12: > http://d.puremagic.com/issues/show_bug.cgi?id=284 > struct A(int x) { } > > template B(uint x) { > static if (x == 3) > static assert(is(A!(3) == A!(x))); > } > > static this() { B!(3); } > > --- > > file.d(5): static assert (is(A == A )) is false > > --- > > Two instances of A!(3) is created. One with an int value type and the other one > with a uint(!) value type. The mangled forms are *__T1AVi3Z1A and *__T1AVk3Z1A > respectively. For the mangling aspect see http://d.puremagic.com/issues/show_bug.cgi?id=109 Thomas -----BEGIN PGP SIGNATURE----- iD8DBQFE4FsmLK5blCcjpWoRAu3yAJ4ovoL9rE78OwWyhWsKze0rHEdC8ACgnrAa LrfyQ2b7PaBNIOj95s13bSQ= =B//3 -----END PGP SIGNATURE-----

The generated template instance is based on the types of the arguments *before* they are implicitly converted to the template parameter type. This is deliberately designed this way so that new template specializations can be added without upsetting existing code. I adjusted the template documentation to clarify this point. Not a bug.