Mathias LANG

Well, the `BlackEdder/ggplotd` failure is interesting. The constructor turns into a copy constructor and that fails with `inout`.

> I think it would be better to change the `error` into a `deprecation` s.t. projects can integrate `-preview=in` more easily. There is two reasons why it's not a deprecation:...

> Agreed but that's an implementation detail(?) (which you disliked IIRC?) Well it does affect tools that do parsing only. And it was made to be consistent with the other...

Current status: # Need tags - `Style` & `dlang-community/D-Scanner`: D-Scanner [needs a new release](https://github.com/dlang-community/D-Scanner/pull/820#issuecomment-691644431). - `dlang/dub`: Compiles and run, but the test-suite depends on pulling old dub versions. I'm just...

@jondegenhardt : Indeed, `in ref` => `const ref` is the path of least resistance. And the only path for libraries. > Switching to `in` by itself doesn't sound like it'll...

> I haven't been able to get clarity on which preview switches are definitely being added to the language and which are experiments/candidates I think all switches but `rvaluerefparams` are...

As replied on the issue, changing `immutable` data is specified as UB currently, and IMHO should stay that way. Having `@system` bypass a `scope` checks is acceptable, but having it...

> @Geod24 half of my PRs don't make it through the test suite with random failures in the test suite. The fix, of course, is to recognize that the failure...

Casting away immutability is fine, it's mutating previously-immutable data that isn't. I'll take a look at `core.lifetime`.

The code was introduced by https://github.com/dlang/druntime/commit/0868bef718432b4deea70c40a81a19143308bb09 so CC @kinke . The commit does mention that this is needed to allow calling differently-qualified copy ctor (e.g. immutable copy ctor). Is that...