Hi! I'm experimenting with swift-testing and Swift 6 strict concurrency checking. I'm seeing some not-super-clear build failure. Maybe this is something that should work and might be fixed later this year… or maybe it's a legit failure and I should rethink how this test is built.
Here is a test that builds (and passes) from Swift 6 strict concurrency:
func doSomething(_: Int) async throws {
}
@Suite actor Tests {
let count = 0
}
extension Tests {
@Test func exampleBuilds() async throws {
let count = 0
try await doSomething(count)
await #expect(throws: Never.self) {
try await doSomething(count)
}
try await doSomething(self.count)
}
}
Here is a test that fails to build from Swift 6 strict concurrency:
extension Tests {
@Test func exampleFails() async throws {
await #expect(throws: Never.self) {
try await doSomething(self.count)
}
}
}
Here is the failure:
24 | await #expect(throws: Never.self) {
25 | try await doSomething(self.count)
26 | }
+--- macro expansion #expect ----------------------------------------
|1 | Testing.__checkClosureCall(throws: Never.self, performing: {
| | `- error: sending 'self'-isolated value of type '() async throws -> ()' with later accesses to nonisolated context risks causing data races
|2 | try await doSomething(self.count)
|3 | }, expression: .__fromSyntaxNode("try await doSomething(self.count)"), comments: [], isRequired: false, sourceLocation: Testing.SourceLocation()).__expected()
+--------------------------------------------------------------------
27 | }
28 | }
Any ideas about what else I can do about that? I don't see this error when building from Swift 5 mode. Is this error legit? Should I rethink how I'm building this test? Any ideas about that would be great. Thanks!