Sema: Relax distributed actor typechecking for swiftinterfaces #65431
Conversation
tshortli
requested review from
hborla,
ktoso,
xedin and
slavapestov
as code owners
|
@swift-ci please test |
| @@ -68,7 +68,6 @@ static VarDecl* | |||
| if (!var->getInterfaceType()->isEqual(expectedType)) | |||
| return nullptr; | |||
|
|
|||
| assert(var->isSynthesized() && "Expected compiler synthesized property"); | |||
| @@ -643,6 +643,12 @@ bool swift::checkDistributedActorProperty(VarDecl *var, bool diagnose) { | |||
| void swift::checkDistributedActorProperties(const NominalTypeDecl *decl) { | |||
| auto &C = decl->getASTContext(); | |||
|
|
|||
| auto sourceFile = decl->getDeclContext()->getParentSourceFile(); | |||
| if (sourceFile && sourceFile->Kind == SourceFileKind::Interface) { | |||
Some decls that are expected to be synthesized for distributed actors are printed explicitly in swiftinterfaces so diagnostics and assertions need to take that possibility into account. Resolves rdar://108533918
tshortli
force-pushed
the
distributed-actor-swiftinterface
branch
from
9577c3c to
0e9b8a1
Compare
|
@swift-ci please smoke test |
Thanks for thinking of that Pavel! I experimented with a few different approaches to seeing if that might be an issue but I wasn't able to reproduce any failures. I tried |
|
That sounds reasonable but @ktoso knows more about the system/id synthesis inside of initializers than I do. |
|
Since this change appears to succeed at unblocking module interface verification I'm going to land this. If we discover ways in which SILGen also needs to be tweaked let's do that in a separate PR. |
|
Hmmm, as far as I can tell these would be the only place where we're diagnosing errors on the synthesized things so this should be enough. Thank you very much for the investigation and fix @tshortli! |
Some decls that are expected to be synthesized for distributed actors are printed explicitly in swiftinterfaces so diagnostics and assertions need to take that possibility into account.
Resolves rdar://108533918