I also added `// skip-codegen` to each one, to address potential concerns
that this change would otherwise slow down our test suite spending time
generating code for files that are really just meant to be checks of
compiler diagnostics.
(However, I will say: My preference is to not use `// skip-codegen` if
one can avoid it. We can use all the testing of how we drive LLVM that
we can get...)
(Updated post rebase.)
This test specifically notes that it does not want to invoke the
linker, due to the way it (IMO weakly) exercises the `#[link=...]`
attribute.
In any case, removing the the `#[rustc_error]` here uncovered an
"invalid windows subsystem" error that was previously not included in
the transcript of diagnostic output. So that's a step forward, (right?).
NLL Diagnostic Review 3: Missing errors for borrows of union fields
Fixes#55675.
This PR modifies a test to make it more robust (it also fixes indentation on a doc comment, but that's not the point of the PR). See the linked issue for details.
r? @pnkfelix
Fix tracking issue numbers for some unstable features
And also remove deprecated unstable `#[panic_implementation]` attribute that was superseded by stable `#[panic_handler]` and doesn't have an open tracking issue.
NLL has increased precision in its analysis of drop order, and we want
the test annotations to deliberately reflect this by having fewer
ERROR annotations for NLL than for AST-borrowck. The best way to get
this effect is via `// revisions`.
As a drive-by, also added uses of all the borrows just to make it
clear that NLL isn't somehow sidestepping things by using shorter
borrows than you might have otherwise expected. (Of course, the added
uses do not make all that much difference since the relevant types all
declare `impl Drop` and thus those drops have implicit uses anyway.)
This is a variant of `ui/borrowck/borrowck-closures-mut-of-imm.rs`
that I used to help identify what changes I needed to make to the
latter file in order to recover its instances of E0524 under NLL.
(Basically this test includes the changes you'd need to make to
`ui/borrowck/borrowck-closures-mut-of-imm.rs` in order to get rid of
occurrences of E0596. And then I realized that one needs to add
invocations of the closures in order to properly extend the mutable
reborrows in a manner such that NLL will roughly match AST-borrowck.)
This is based on the feedback from estebank:
"""
I believe that test can be removed outright. It'd be impossible for a
new change to go through that breaks this kind of output without it
being picked up by multiple other `stderr` tests. This is an artifact
of the transition period to the "new" output style.
"""
see: https://github.com/rust-lang/rust/issues/52663#issuecomment-422155551
This is not strictly necessary to make this test "more robust with
respect to NLL"; its just an attempt to narrow the scope of the test
and focus on its core.
Most of the time we want to robustify tests, but in this case this
test is deliberately encoding artifacts of AST-borrowck. So instead
of adding artificial uses that would obscure the aspects of
AST-borrowck that are being tests, we instead use revisions and then
mark the cases that apply to NLL as well as AST-borrowck.
The error message is sub-par, but fixing that requries moving ScalarMaybeUndef
to librustc which would conflict badly with another PR that is in flight.
In each of the three cases in this test, there is a mutable borrow
of some field of the union and then a shared borrow of some other field
immediately following.
Under NLL, the mutable borrow is killed straight away as it isn't
used later - therefore not causing a conflict with the shared borrow.
This commit adds a use of the first mutable borrow to force the intended
errors to appear under NLL.
Take 2: Implement object-safety and dynamic dispatch for arbitrary_self_types
This replaces #50173. Over the months that that PR was open, we made a lot of changes to the way this was going to be implemented, and the long, meandering comment thread and commit history would have been confusing to people reading it in the future. So I decided to package everything up with new, straighforward commits and open a new PR.
Here are the main points. Please read the commit messages for details.
- To simplify codegen, we only support receivers that have the ABI of a pointer. That means they are builtin pointer types, or newtypes thereof.
- We introduce a new trait: `DispatchFromDyn<T>`, similar to `CoerceUnsized<T>`. `DispatchFromDyn` has extra requirements that `CoerceUnsized` does not: when you implement `DispatchFromDyn` for a struct, there cannot be any extra fields besides the field being coerced and `PhantomData` fields. This ensures that the struct's ABI is the same as a pointer.
- For a method's receiver (e.g. `self: Rc<Self>`) to be object-safe, it needs to have the following property:
- let `DynReceiver` be the receiver when `Self = dyn Trait`
- let `ConcreteReceiver` be the receiver when `Self = T`, where `T` is some unknown `Sized` type that implements `Trait`, and is the erased type of the trait object.
- `ConcreteReceiver` must implement `DispatchFromDyn<DynReceiver>`
In the case of `Rc<Self>`, this requires `Rc<T>: DispatchFromDyn<Rc<dyn Trait>>`
These rules are explained more thoroughly in the doc comment on `receiver_is_dispatchable` in object_safety.rs.
r? @nikomatsakis and @eddyb
cc @arielb1 @cramertj @withoutboats
Special thanks to @nikomatsakis for getting me un-stuck when implementing the object-safety checks, and @eddyb for helping with the codegen parts.
EDIT 2018-11-01: updated because CoerceSized has been replaced with DispatchFromDyn
rustc: improve E0669 span
E0669 refers to an operand that cannot be coerced into a single LLVM
value, unfortunately right now this uses the Span for the entire inline
assembly statement, which is less than ideal.
This commit preserves the Span from HIR, which lets us emit the error
using the Span for the operand itself in MIR.
r? @nagisa
cc/ @parched
