Previously, if a request R evaluated itself N times, we would emit N
"circular reference" diagnostics. These add no value, so instead let's
cache the user-provided default value on the first circular evaluation.
This changes things slightly so that instead of returning an
llvm::Expected<Request::OutputType>, various evaluator methods take
a callback which can produce the default value.
The existing evaluateOrDefault() interface is unchanged, and a new
evaluateOrFatal() entry point replaces
llvm::cantFail(ctx.evaluator(...)).
Direct callers of the evaluator's operator() were updated to pass in
the callback. The benefit of the callback over evaluateOrDefault() is
that if the default value is expensive to constuct, like a dummy
generic signature, we will only construct it in the case where a
cycle actually happened, otherwise we just delete the callback.
(cherry picked from commit b8fcf1c709efa6cd28e1217bd0efe876f7c0d2b7)
In some cases we get a `ConstraintKind::SelfObjectOfProtocol` here.
That's typically for an existential cast. In such a case, we actually
get a better standard diagnostic than what MustBeCopyable can produce
currently. So this change narrows things to what MustBeCopyable is
better at diagnosing.
Unify with `CTP_ReturnStmt`, and have the
SyntacticElementTarget carry the ReturnStmt for
regular type-checking, which we can use to record
implied returns.
Track the implied result exprs in the constraint
system, and allow arbitrary propagation of
implied results down if/switch expression
branches. This is required for allowing implied
results in non-single-expression closures.
Remove this bit from function decls and closures.
Instead, for closures, infer it from the presence
of a single return or single expression AST node
in the body, which ought to be equivalent, and
automatically takes result builders into
consideration. We can also completely drop this
query from AbstractFunctionDecl, replacing it
instead with a bit on ReturnStmt.
Not quite NFC because apparently the representation bleeds into what's
accepted in some situations where we're supposed to be warning about
conflicts and then making an arbitrary choice. But what we're doing
is nonsense, so we definitely need to break behavior here.
This is setting up for isolated(any) and isolated(caller). I tried
to keep that out of the patch as much as possible, though.
Use an optional isolated parameter to this new `next(_:)` overload to
keep it on the same actor as the caller, and pass `#isolation` when
desugaring the async for..in loop. This keeps async iteration loops on
the same actor, allowing non-Sendable values to be used with many
async sequences.
This couples together several changes to move entirely from
`@rethrows` over to typed throws:
* Use the `Failure` type to determine whether an async for-each loop
will throw, rather than depending on rethrows checking
* Introduce a special carve-out for `rethrows` functions that have a
generic requirement on an `AsyncSequence` or `AsyncIteratorProtocol`,
which uses that requirement's `Failure` type as potentially being part
of the thrown error type. This allows existing generic functions like
the following to continue to work:
func f<S: AsyncSequence>(_: S) rethrows
* Switch SIL generation for the async for-each loop from the prior
`next()` over to the typed-throws version `_nextElement`.
* Remove `@rethrows` from `AsyncSequence` and `AsyncIteratorProtocol`
entirely. We are now fully dependent on typed throws.
It should be the responsibility of callers to map the type to a
contextual type, as needed. When it's not possible or repetitive to do
so, there is a special-purpose function `isInterfaceTypeNoncopyable` for
Sema.
Remove keypath subtype asserts; always use cached root type
Add tests for keypaths converted to funcs with inout param
Add unit test for overload selection
Handle requirement failures in result builder `build*` methods explicitly
and give them a high impact rating because issues if such positions
imply that the transform didn't work and it shouldn't shadow errors
in user code.
Resolves: rdar://111120803
Resolves: rdar://120342129
Doing so improves performance if closure is passed as an argument
to a (heavily) overloaded declaration, avoid unrelated errors,
propagate holes, and record a more impactful fix.
Start classifying all potential throw sites within a constraint
system and associate them with the nearest enclosing catch node. Then,
determine the thrown error type for a given catch node by taking the
union of the thrown errors at each potential throw site. Use this to
compute the error type thrown from the body of a `do..catch` block
within a closure.
This behavior is limited to the upcoming feature `FullTypedThrows`.
The errorUnion type operation specifies how thrown error types are
combined when multiple errors are thrown in the same context. When
thrown error types can have type variables in them, we sometimes cannot
resolve the errorUnion until the type variables have substitutions. In
such cases, we need to persist the result of errorUnion in the
constraint solver.
Introduce the ErrorUnionType to do exactly that, and update the core
errorUnion operation to produce an ErrorUnionType when needed. At
present, this code is inert, because any errorUnion operation today
involves only concrete types. However, inference of thrown errors in
closures will introduce type variables, and depend on this.
We already need to track the inverses separate from the members in a
ProtocolCompositionType, since inverses aren't real types. Thus, the
only purpose being served by InverseType is to be eliminated by
RequirementLowering when it appears in a conformance requirement.
Instead, we introduce separate type InverseRequirement just to keep
track of which inverses we encounter to facilitate cancelling-out
defaults and ensuring that the inverses are respected after running
the RequirementMachine.
