Bug 19060 – [REG2.081] Incorrect "Using this as a type is deprecated" error

Testcase: ``` struct S { bool x; public template yoyo() { alias yoyo = this.x; // line 5 } public ref foo() { return yoyo!(); } } ``` Works correctly with dmd < 2.081, fails (incorrect deprecation) with 2.081: ❯ dmd testcase.d -de testcase.d(5): Deprecation: Using this as a type is deprecated. Use typeof(this) instead

This was changed on purpose. https://dlang.org/changelog/2.081.0.html#deprecate_this_super_as_types

Okay looks like I was wrong. Removing the class it would be equivalent at the module level to: --- module self; bool x; template Foo() { alias y = self.x; } alias Bar = Foo!(); ---

As far as I can see, the deprecation is correct. You can't get an alias of an instance field. The alias was to `S.x` the whole time. You're just being forced to spell it out now.

Indeed, this is not a regression, it is intended behavior.

OK, then this is still broken and should receive a deprecation message too: ``` struct S { int a; bool x; public ref foo() { alias yoyo = this.x; return yoyo; } } ```

(In reply to johanengelen from comment #5) > OK, then this is still broken and should receive a deprecation message too: > > ``` > struct S { > int a; > bool x; > public ref foo() { > alias yoyo = this.x; > return yoyo; > } > } > ``` Actually, I don't think it should. Foo can only be called on an instance, therefore `this` does make sense in that context. Imagine that you would define a local variable x, if you wouldn't be able to use `this` then the local would mask the member everywhere. I suggest to close this as invalid.

(In reply to RazvanN from comment #6) > (In reply to johanengelen from comment #5) > > OK, then this is still broken and should receive a deprecation message too: > > > > ``` > > struct S { > > int a; > > bool x; > > public ref foo() { > > alias yoyo = this.x; > > return yoyo; > > } > > } > > ``` > > Actually, I don't think it should. Foo can only be called on an instance, > therefore `this` does make sense in that context. The alias is still to `S.x`. It still doesn't carry `this`. For example, this might be surprising: ---- struct S { int x = 1; int foo(S other) { alias yoyo = other.x; return yoyo; } } void main() { import std.stdio; auto s1 = S(1); auto s2 = S(2); writeln(s1.foo(s2)); /* prints "1" */ } ---- From the alias declaration, one might expect to get `other.x`, i.e. 2. > Imagine that you would > define a local variable x, if you wouldn't be able to use `this` then the > local would mask the member everywhere. You can still write `alias x = S.x;`.

(In reply to ag0aep6g from comment #7) > (In reply to RazvanN from comment #6) > > (In reply to johanengelen from comment #5) > > > OK, then this is still broken and should receive a deprecation message too: > > > > > > ``` > > > struct S { > > > int a; > > > bool x; > > > public ref foo() { > > > alias yoyo = this.x; > > > return yoyo; > > > } > > > } > > > ``` > > > > Actually, I don't think it should. Foo can only be called on an instance, > > therefore `this` does make sense in that context. > > The alias is still to `S.x`. It still doesn't carry `this`. > > For example, this might be surprising: > > ---- > struct S { > int x = 1; > int foo(S other) > { > alias yoyo = other.x; > return yoyo; > } > } > > void main() > { > import std.stdio; > auto s1 = S(1); > auto s2 = S(2); > writeln(s1.foo(s2)); /* prints "1" */ > } > ---- > > From the alias declaration, one might expect to get `other.x`, i.e. 2. > > > > Imagine that you would > > define a local variable x, if you wouldn't be able to use `this` then the > > local would mask the member everywhere. > > You can still write `alias x = S.x;`. This is surprising, indeed, and I would say that this is a bug. Why would it do that?

(In reply to RazvanN from comment #8) > This is surprising, indeed, and I would say that this is a bug. Why would it > do that? Because `alias yoyo = other.x;` is the same as `alias yoyo = this.x;`. They both mean `alias yoyo = S.x;`. An alias can't refer to a field of a specific instance.

(In reply to ag0aep6g from comment #9) > Because `alias yoyo = other.x;` is the same as `alias yoyo = this.x;`. They > both mean `alias yoyo = S.x;`. An alias can't refer to a field of a specific > instance. I believe that sums up this issue, and I'll mark it as invalid.