Closed
Description
The current rules are not really sound. It will permit e.g. an &mut
pointer to be created for a mut
field found in an &T
, which is aliasable. This is bad because the &mut
pointer may not be the only pointer to that data.
The rewritten version I am trying to land as part of #5074 includes a fix but it's commented out because it causes compile errors in io.rs. Fixing those errors requires cleaning up object types and converting things to &mut self
and I didn't want to block #5074 on those changes. (Hence this follow-on bug)