NLL elides type behind a borrow (and its Copy-ness) when reporting move error

[pnkfelix]

If you look at the AST-borrowck output for a case like borrowck-vec-pattern-nesting.rs, you see output like this:

rust/src/test/ui/borrowck/borrowck-vec-pattern-nesting.stderr

Line 19 in 20af72b

error[E0508]: cannot move out of type `[std::boxed::Box<isize>]`, a non-copy slice

but NLL only reports this:

rust/src/test/ui/borrowck/borrowck-vec-pattern-nesting.nll.stderr

Line 19 in 20af72b

error[E0507]: cannot move out of borrowed content

It is probably useful to include the type behind the borrow (and the fact that it is non-copy) when that is relevant to the move being attempted.

[pnkfelix]

Hmm, interesting; it seems like the MIR dataflow is at least making an attempt to capture the relevant distinctions here, at least based on the code at the following links:

rust/src/librustc_mir/dataflow/move_paths/builder.rs

Lines 148 to 156 in 2b5aa90

	ty::TySlice(_) =>
	return Err(MoveError::cannot_move_out_of(
	mir.source_info(self.loc).span,
	InteriorOfSliceOrArray {
	ty: place_ty, is_index: match proj.elem {
	ProjectionElem::Index(..) => true,
	_ => false
	},
	})),

rust/src/librustc_mir/borrow_check/mod.rs

Lines 141 to 143 in 2b5aa90

	IllegalMoveOriginKind::InteriorOfSliceOrArray { ty, is_index } => {
	tcx.cannot_move_out_of_interior_noncopy(span, ty, is_index, origin)
	}

So... is that we never actually construct instances of such errors when doing dataflow for MIR-borrowck? (Or is there some deeper issue with the error reporting?) I will look a bit further into this now.

[pnkfelix]

I think I have this figured out. I hope to have a PR up later today.