[Runtime] Fix crash in protocol conformance checks on KVO artificial subclasses. #37973
Conversation
…subclasses. tryGetCompleteMetadataNonblocking crashes on artificial subclasses due to the NULL type descriptor. Explicitly check for artificial subclasses in getSuperclassForMaybeIncompleteMetadata and immediately return their Superclass field. Artificial subclasses are always fully initialized so we don't need to do anything special for them. rdar://72583931
| // tryGetCompleteMetadataNonblocking will crash on them. However, they're | ||
| // always fully set up, so we can just skip it and fetch the Subclass field. | ||
| if (classMetadata->isTypeMetadata() && classMetadata->isArtificialSubclass()) | ||
| return {classMetadata->Superclass, MetadataState::Complete}; |
There was a problem hiding this comment.
In other places, we do this check for artificial subclasses in a loop, crawling up the superclass tree until we find a real class. I don't know if that's just paranoia, though.
There was a problem hiding this comment.
That can happen at least in theory. No idea if it happens in reality, but it's good to handle. In this particular case, the loop is exterior to this function, as other code uses this function to walk the entire superclass chain. Returning artificial subclasses in that loop is unnecessary but harmless.
|
@swift-ci please test |
|
Build failed |
|
Build failed |
|
@swift-ci please test macOS platform |
|
Build failed |
|
@swift-ci please test macos platform |
|
Build failed |
|
@swift-ci please smoke test macOS platform |
tryGetCompleteMetadataNonblocking crashes on artificial subclasses due to the NULL type descriptor. Explicitly check for artificial subclasses in getSuperclassForMaybeIncompleteMetadata and immediately return their Superclass field. Artificial subclasses are always fully initialized so we don't need to do anything special for them.
rdar://72583931