This is a narrow workaround for a regression from
9e3d0e0a8c.
There is no reason to skip this logic for protocol extension
members, except that doing so happens to break existential
opening in an expression that involves a call to Array.init
elsewhere.
However there is an underlying issue here with existential opening,
which doesn't seem to work right in the presence of overloading.
The test case demonstrates the fixed problem, together with an
existing bug that points to the underlying problem.
Fixes rdar://160389221.
We allow placeholder types in interface types in certain cases to allow
better recovery since we can suggest the inferred type as a replacement.
When referencing those decls though we need to make sure we record a fix
since we cannot form a valid solution with them.
`KeyPath` types now have conformance requirements placed on their
`Root` and `Value` types which need to be checked even when there
is a key path to function conversion involved, otherwise the solver
would be accepting invalid code.
Note that without function conversion the requirements come from
a type opened during assignment - https://github.com/swiftlang/swift/pull/80081/files.
Resolves: https://github.com/swiftlang/swift/issues/84150
Eagerly bind invalid type references to holes and propagate contextual
type holes in `repairFailures`. This avoids some unnecessary diagnostics
in cases where we have an invalid contextual type.
If the argument type is an array and it's passed to an imported declaration
that accepts a raw pointer, the solver should use an "array-to-c-pointer"
conversion instead of the one for pointers.
Resolves: rdar://158629300
For a method key path use the locator for the apply itself rather
than the member, ensuring we handle invalid cases where the apply is
the first component, and providing more accurate location info.
If the base type is composed with marker protocol(s) i.e.
`<<Type>> & Sendable`, let's skip this check because such
compositions are always opened and simplified down to a
superclass bound post-Sema.
Resolves: rdar://148782046
Set an upper bound on the number of chained lookups we attempt to
avoid spinning while trying to recursively apply the same dynamic
member lookup to itself.
rdar://157288911
The current implementation of the check accounts only for the overload
choices present in the initial lookup but for some situations, like
bridged or optional base types, `performMemberLookup` uses a secondary
lookup as well, results of which are ignored.
Let's fold the check into `addChoice` instead and set the the flag there
to make sure that all of the choices are considered.
Resolves: rdar://143586718
If result of `CGFloat` -> `Double` conversion is injected into an optional
it should be ranked based on depth just like when locator is fully simplified.
For example:
```swift
func test(v: CGFloat?) {
_ = v ?? 2.0 / 3.0
}
```
In this expression division should be performed on `Double` and result
narrowed down (based on the rule that narrowing conversion should always
be delayed) but `Double` -> `CGFloat?` was given an incorrect score and
instead of picking `?? (_: T?, _: T) -> T` overload, the solver would
use `?? (_: T?, _: T?) -> T?`.
(cherry picked from commit cb876cbd9e)
Follow-up for https://github.com/swiftlang/swift/pull/82326.
The optional injection is only viable is the wrapped type is
not yet resolved, otherwise it's safe to wrap the optional.
We need to be very careful while matching types to test whether a
fix is applicable or not to avoid adding extraneous fixes and failing
the path early. This is a temporary workaround, the real fix would
be to let `matchTypes` to propagate `TMF_ApplyingFixes` down.
Resolves: rdar://154010220
Resolves: https://github.com/swiftlang/swift/issues/82397
If we have a tuple with unresolved pack expansion on one side
and an optional type on the other, prevent `matchTypes` from
wrapping optional into a one-element tuple because the matching
should be handled as part of the optional injection.
Resolves: rdar://152940244
Package the flag into `performanceHacksEnabled()` method on
`ConstraintSystem` and start using it to wrap all of the hacks
in constraint generator and the solver.
Escaping solver-allocated types into a nested allocation arena is
problematic since we can e.g lazily compute the `ContextSubMap` for a
`NominalOrBoundGenericNominalType`, which is then destroyed when we
exit the nested arena. Ensure we don't pass any types with type
variables or placeholders to `typesSatisfyConstraint`.
rdar://152763265
Fixes a crash on invalid. The previous logic was causing a label
mismatch constraint fix to be recorded for an unlabeled trailing closure
argument matching a variadic paramater after a late recovery argument
claim in `matchCallArgumentsImpl`, because the recovery claiming skips
arguments matching defaulted parameters, but not variadic ones. We may
want to reconsider that last part, but currently it regresses the
quality of some diagnostics, and this is a targeted fix.
The previous behavior is fine because the diagnosis routine associate
with the constraint fix (`diagnoseArgumentLabelError`) skips unlabeled
trailing closures when tallying labeling issues — *unless* there are no
other issues and the tally is zero, which we assert it is not.
Fixes rdar://152313388.
- Mismatch in tuple element position should reference whole tuple
with a note for mismatch position;
- Situations where optional object type is not a class but matched
against `AnyObject` have a tailored fix.
Attempting to propagate generic argument failures up is not always
reliable, `matchDeepEqualityTypes` should avoid using `TMF_ApplyingFix`
while dealing with optionals and instead let `repairFailures` decide
whether to use generic arguments mismatch fix to a more general one.
If generic arguments mismatch ends up being recorded on the result
of the chain or `try` expression it means that there is a contextual
conversion mismatch.
For optional conversions the solver currently generates a disjunction
with two choices - bind and optional-to-optional conversion which is
anchored on the contextual expression. If we can get a fix recorded
there that would result in a better diagnostic. It's only possible
for optional-to-optional choice because it doesn't bind the
variable immediately, so we need to downgrade direct fixes to prevent
`bind` choice from considered better.
