This fails to compile ------ deprecated(""~"") module a; ------
Although the OP code is legitimate from the D grammar spec, there's not yet proper definition/implementation for the semantic analysis. I'm not sure how compiler should behave with: deprecated(a.foo() ~ a.bar()) module a; string foo() { /* some CTFEable code 1 */ } string bar() { /* some CTFEable code 2 */ } ---- Note that if we replace module with struct: deprecated(S.foo) struct S // line 1 { static foo() { return "hello"; } } alias P = S*; // line 5 Current compiler prints: test.d(1): Deprecation: struct test.S is deprecated test.d(5): Deprecation: struct test.S is deprecated - hello The use of S in custom message expression exposes incompletion of semantic analysis process. It's bad behavior.
It could at least be made to work with string literals and manifest constants, right?
(In reply to Jack Stouffer from comment #2) > It could at least be made to work with string literals and manifest > constants, right? I think it's not good to implement vulnerable behavior. It would just bloat compiler maintenance cost. By the way, do you have a concrete situation you would like to use contatenation, manifest constant, or other compile time constructed message for module deprecation?
One could simply put pragma message in that module. module a; pragma(msg, __FILE__,"(",__LINE__,"): this module is deprecated");
(In reply to Kenji Hara from comment #3) > (In reply to Jack Stouffer from comment #2) > > It could at least be made to work with string literals and manifest > > constants, right? > > I think it's not good to implement vulnerable behavior. It would just bloat > compiler maintenance cost. But this is already the behavior for everything but module declarations: ----- struct A { deprecated(""~"") int a; } void main() { A s; s.a = 4; } ----- $ dmd -run test.d Deprecation: variable test.A.a is deprecated - > By the way, do you have a concrete situation you would like to use > contatenation, manifest constant, or other compile time constructed message > for module deprecation? Yes, in Phobos we are trying to enforce Phobos style guides via dscanner, and we need this in order to make every line in Phobos less than 120 characters.
Sometime I look at the timing, and I think we're doing a really bad job at marketing our awesome features. The fix for this was merged in February 2016, so 3 months before this bug report was open. Fixed by: https://github.com/dlang/dmd/pull/5302 Marking as duplicate of the original issue. *** This issue has been marked as a duplicate of issue 12954 ***