We're currently building a toolchain so that people could try this out in action. I'll post it here once the build completes.
Toolchains:
- Ubuntu 20.04 (x86_64)
- macOS - pending
The implementation in the preview toolchain coming up is not compatible with -enable-actor-data-race-checks when configured to "warn" mode, and it will crash even if asked to just log warnings. We are working on addressing this limitation soon.
I would also like to point out the relationship of this proposal and the ongoing review of dynamic enforcement: SE-0423: Dynamic actor isolation enforcement from non-strict-concurrency contexts
The implicitly inserted checks discussed in that proposal, would also benefit from this improved checking semantics. This might be most interesting for the MainActor, where main actor functions are called from "raw Main queue", as discussed in this proposal.