When defaulting to main-actor isolation, types that have synthesized
conformances (e.g., for Equatable, Hashable, Codable) were getting
nonisolated members by default. That would cause compiler errors
because the conformances themselves defaulted to main-actor isolation
when their types were.
Be careful to only mark these members as 'nonisolated' when it makes
sense, and leave them to get the isolation of their enclosing type
when the conformance might have isolation. This ensures that one can
use synthesis of these protocols along with default main-actor mode.
There is a one-off trick here to force the synthesized CodingKeys to
be nonisolated, because the CodingKey protocol requires Sendable.
We'll separately consider whether to generalize this rule.
More of rdar://150691429.
Cherry-pick of #80547 for the 6.2 release branch.
---
Explanation: This cherry picks the implementation of SE-0477 to add a
string interpolation method with a `default:` parameter for optional
interpolation values.
Main Branch PR: https://github.com/swiftlang/swift/pull/80547
Risk: Low.
Reviewed By: @stephentyrone
Resolves: rdar://150865613
Testing: New tests for the string interpolations and fix-its.
It has indirect effects on the accessors, so it shouldn’t matter, but we can defensively redirect the query to the API counterpart anyway.
This was the last `InferredInABIAttr` attribute, so we can now remove all of the infrastructure involved in supporting attribute inference.
CustomAttr backs four different features, each of which requires a different behavior in `@abi`:
• Global actors: Permitted (and permitted to vary) since they can affect mangling
• Result builders: Forbidden inside an `@abi` since they have no ABI impact
• Property wrappers: Forbidden both inside an `@abi` and on a decl with an `@abi` since it’s not clear how we would apply `@abi` to the auxiliary decls
• Attached macros: Forbidden inside an `@abi` since an ABI-only decl has no body, accessors, members, peers, extensions, or (currently) conformances
Implement these behaviors (outside of `ABIDeclChecker` since they can’t be described there).
Macro-related tests are not included in this commit; they require matching swift-syntax changes which are being negotiated.
Macro expansions are now treated like a part of the source file they belong to, for purposes of the “second declaration is the one that’s diagnosed” rule. This helps stabilize a behavior that was easy to perturb.
The decl checker was effectively not being run on these because we weren’t typechecking the PBD and typechecking the VarDecl itself is basically a no-op.
Some notes:
1. In most cases, I think we were getting lucky with this by just inferring the
closure's isolation from its decl context. In the specific case that we were
looking at here, this was not true since we are returning from an @concurrent
async function a nonisolated(nonsending) method that closes over self. This
occurs since even when NonisolatedNonsendingByDefault we want to start importing
objc async functions as nonisolated(nonsending).
2. I also discovered that in the ActorIsolationChecker we were not visiting the
inner autoclosure meaning that we never set the ActorIsolation field on the
closure. After some discussion with @xedin about potentially visiting the
function in the ActorIsolationChecker, we came to the conclusion that this was
likely to result in source stability changes. So we put in a targeted fix just
for autoclosures in this specific case by setting their actor isolation in the
type checker.
3. Beyond adding tests to objc_async_from_swift to make sure that when
NonisolatedNonsendingByDefault is disabled we do the right thing, I noticed that
we did not have any tests that actually tested the behavior around
objc_async_from_swift when NonisolatedNonsendingByDefault is enabled. So I added
the relevant test lines so we can be sure that we get correct behavior in such a
case.
rdar://150209093
(cherry picked from commit ced96aa5cd)
For macro definition checking, we use the range of the `macro`
declaration and re-parse it with `SwiftParser`. Previously it uses the
range including the attributes, but that can result invalid code because
the attribute can be in a `#if ... #endif` region.
Since we don't use attributes for checking the definition, just use the
range without the attributes instead.
rdar://150805795
(cherry picked from commit 8519e71602)
Static member referenced were marked as `@Sendable` by `InferSendableFromCaptures`
because metatypes used to be always Sendable which is no longer the case, so in
order to maintain the source compatibility we need to downgrade missing `@Sendable`
to a warning for unapplied static member references.
This affects primarily operators at the moment because other static members
form a curry thunk with a call inside and would be diagnosed as a capture.
Resolves: rdar://150777469
(cherry picked from commit a57310b61d)
This request was looking through to the root conformance, which could
mess with the caching bits. Sink the "is nonisolated conformance" bit
down into ProtocolConformance, and have the request for a non-root
conformance be defined in terms of the request for the root
conformance.
If all of the witnesses to a conformance are nonisolated, then infer that
conformance as nonisolated rather than global-actor-isolated. This is
only relevant when InferIsolatedConformances is enabled, and prevents
that inference to help maintain source compatibility.
We can always get it back from the git history.
rdar://150695113
(cherry picked from commit 9d59dbed17)
Conflicts:
include/swift/AST/DiagnosticsSema.def
include/swift/Basic/Features.def
test/abi/macOS/arm64/concurrency.swift
test/abi/macOS/x86_64/concurrency.swift
per SE-0431, function conversions from an @isolated(any) function to a synchronous,
non-@isolated(any) function type should not be allowed. this adds a warning during
type checking to enforce this, which will be an error in a future major
language mode.
(cherry picked from commit 629e208970)
When `NonisolatedNonsendingByDefault` is enabled it should infer
`nonisolated(nonsending)` for both async functions and storage as
specified by the proposal.
(cherry picked from commit 8a9bbd5114)
Otherwise, we break memory safety by accessing a generic parameter that does not
exist resulting in undefined behavior.
rdar://150351351
(cherry picked from commit fb3c710212)
It is possible for a C++ class template to inherit from a specialization
of itself. Normally, these are imported to Swift as separate (unrelated)
types, but when symbolic import is enabled, unspecialized templates are
imported in place of their specializations, leading to circularly
inheriting classes to seemingly inherit from themselves.
This patch adds a check to guard against the most common case of
circular inheritance, when a class template directly inherits from
itself. This pattern appears in a recent version of libc++,
necessitating this patch. However, the solution here is imperfect as it
does not handle more complex/contrived circular inheritance patterns.
This patch also adds a test case exercising this pattern. The
-index-store-path flag causes swift-frontend to index the C++ module
with symbolic import enabled, without the fix in this patch, that test
triggers an assertion failure due to the circular reference (and can
infinitely recurse in the StorageVisitor when assertions are disabled).
rdar://148026461
(cherry picked from commit 1f2107f357)
Suppose protocol P has a primary associated type A, and we have
a `any P<S>` value. We form the generalization signature <T>
with substitution map {T := S}, and the existential signature
<T, Self where T == Self.A>.
Now, if we call a protocol requirement that takes Self.A.A.A,
we see this is fixed concrete type, because the reduced type of
Self.A.A.A is T.A.A in the existential signature.
However, this type parameter is not formed from the
conformance requirements of the generalization signature
(there aren't any), so we cannot directly apply the outer
substitution map.
Instead, change the outer substitution conformance lookup
callback to check if the reduced type parameter is valid
in the generalization signature, and not just rooted in a
generic parameter of the generalization signature.
If it isn't, fall back to global conformance lookup.
A better fix would introduce new requirements into the
generalization signature to handle this, or store them
separately in the generic environment itself. But this is fine
for now.
- Fixes https://github.com/swiftlang/swift/issues/79763.
- Fixes rdar://problem/146111083.
Similar to what we do for 'throws' checking, perform argument-specific
checking for unsafe call arguments. This provides more detailed failures:
```
example.swift:18:3: warning: expression uses unsafe constructs but is not
marked with 'unsafe' [#StrictMemorySafety]
16 | x.f(a: 0, b: 17, c: nil)
17 |
18 | x.f(a: 0, b: 17, c: &i)
| | `- note: argument 'c' in call to instance
method 'f' has unsafe type 'UnsafePointer<Int>?'
| `- warning: expression uses unsafe constructs but is not marked
with 'unsafe' [#StrictMemorySafety]
19 | unsafeF()
20 | }
```
It also means that we won't complain for `nil` or `Optional.none`
arguments passed to unsafe types, which eliminates some false
positives, and won't complain about unsafe result types when there is
a call---because we'd still get complaints later about the
actually-unsafe bit, which is using those results.
Fixes rdar://149629670.
(cherry picked from commit ee9487b86f)
When we look through a "self" application while decomposing a function
call in effects checking, keep track of whether we did so. Use this to
ensure that we get the right parameter types when looking at the
"self" application vs. the outer call.
This has always been wrong, but the async/throws effects don't ever
matter for the innermost application (i.e., for `x.m`), so the
inconsistency didn't come up.
(cherry picked from commit 0b264e2d55)
`getType` here can return a null type if the queried expression isn't
part of the solution, which can currently happen for code completion.
I'm working on a more principled fix for this, but until then this is
a low-risk fix that will unblock the stress tester and can be
cherry-picked to 6.2.
rdar://149759542
In b30006837e, I changed the `if`
condition here to check for the absence of type variables as well
as type parameters. This is incorrect; the type variables come up
in ValueWitnessRequest, and the type parameters come up in
associated type inference. We want the matching to be more lax
in the former case.
Fixes rdar://149438520.
In strict concurrency mode some calls could reference a declaration that
is wrapped in one or more function conversion expressions to apply
concurrency related attributes or erase them (such as `@Sendable` or
`@MainActor`). This shouldn't impact constness checking and the checker
should look through such conversions.
Resolves: rdar://148168219
(cherry picked from commit b484e9645d)
When `MemberImportVisibility` is enabled and a declaration from a cross import
overlay is diagnosed because it has not been imported, suggest imports of the
declaring and bystanding modules instead of the cross import overlay module
(which is an implementation detail).
Resolves rdar://149307959.
This is behavior pre-SE-0461 which is the safest option instead
of inferring `nonisolated(nonsending)` and changing the calling
convention and foregoing Sendable checking.
(cherry picked from commit c25d2fab50)
