Places based on misaligned pointers: also mention 'static's as a possible base for place projections

[RalfJung]

In MIR, locals are the only place "roots" since statics get turned into *PTR where PTR is a constant pointer pointing to that static... but in the surface language, STATIC.foo.bar arguably has STATIC as its root and contains no *, so let's add that as a possible case here.

[ehuss]

@RalfJung Just curious, this came up in discussing #1610 where it seems like statics get converted to implicit dereferences as you mentioned. Is that something we need to explicitly say somewhere? Is this a property that should be stated, or is it just a consequence of how MIR works and there could be other ways it was implemented? Is there existing terminology for what this static translation is called?

[RalfJung]

statics get converted to implicit dereferences as you mentioned. Is that something we need to explicitly say somewhere? Is this a property that should be stated, or is it just a consequence of how MIR works and there could be other ways it was implemented?

I consider this to be a MIR implement detail that should not be mentioned in the reference.

Cc @rust-lang/opsem

[digama0]

Agreed that from a language perspective we should treat locals and statics on similar footing and not expose the pointer desugaring.