Bug 1703 – invariant(Class) does not work as expected

Hi, According to the docs I would have expected the following to compile using dmd-2.008 class C { int x; } void main() { invariant(C) c; c = new C; // (1) ok } However, dmd tells me that test.d(9): Error: cannot implicitly convert expression (new C) of type test.C to invariant(C) Also, am I assuming correctly that the following should also compile? If not, how else would one initialize the class? class C { this(int x) { x_ = x; } int x_; } void main() { invariant(C) c; c = new C(10); // ok ?? } Thanks, Markus

On 12/1/07, [email protected] <[email protected]> wrote: > According to the docs I would have expected the following to > compile using dmd-2.008 > > class C > { > int x; > } > > void main() > { > invariant(C) c; > c = new C; // (1) ok > } > > However, dmd tells me that > > test.d(9): Error: cannot implicitly convert expression (new C) of type test.C > to invariant(C) I believe the compiler is correct. Mutable data will not implicitly convert to invariant data. For that you need an explicit cast (new always creates mutable data). There is a template called assumeUnique in std.contracts, but I think it only works for arrays. In any case, all it does is insert a cast, so you could just do that manually, as in: c = cast(invariant(C)) new C; (You could always ask for a language feature inew :) )

Hi Janice, Thanks much for your comments and casting works fine indeed. The main reason I brought this up was that the docs at http://www.digitalmars.com/d/const3.html explicitly state that what I posted should be possible (the snippet of code is pretty much verbatim taken from there). and works fine for structs but not for classes. Hopefully, I didn't completely misunderstand the docs. cheers, Markus

The docs have been fixed.