Restrict static mut

[gereeter]

This is an alternative to #177.

[sfackler]

How are you defining "mutation"? Calling MY_ATOMIC_INT.store(1, SeqCst) certainly modifies it.

[gereeter]

I should have been more clear - I meant direct mutation, without going through UnsafeCell.

[reem]

I'm unclear as to what "directly" mutating something means. Could you clarify this further? Code examples would really help.

[bharrisau]

The only issue I have with this is that static effectively becomes const and static mut becomes static. By that, I mean the only distinction is the location in the binary; rodata or data.

Are there any use cases of the current static mut that can't be handed by UnsafeCell? (UnsafeCell can return a *mut for anyone unfamiliar).

[lilyball]

Am I correct in reading this as static mut now differs from static purely in that static mut may contain UnsafeCell?

I'm still in favor of @nikomatsakis's approach that was described in a comment on #177 (possibly modified, as described in the comments after, to remove static mut in favor of static + UnsafeCell).

[Tobba]

-1
This isnt even an alternative

[nikomatsakis]

Hello! Thank you for the RFC submission. In triage meeting, we decided to close this RFC in favor of #246, as that design encompasses the goals of this one (providing a safe path for people to use Atomic integers and so forth).