Safe code relied upon by unsafe is effectively unsafe

I missed that comment initially. That's almost what the robust keyword is about, see the unsafe mental model linked above. The difference is that robust is more fine grained. Only the Robustness section of the documentation can be assumed to be correct to prove soundness. Libraries may choose to specify their behavior further without having to be accountable for it, and that would just be normal documentation.

1 Like

That could work but it

Aka stability attributes outside of stdlibs

This topic was automatically closed 540 days after the last reply. New replies are no longer allowed.