Mark enums as Sendable #165
Conversation
tristanlabelle
force-pushed
the
tristan/make-enums-sendable
branch
from
82ca079
to
842cc21
Compare
|
Retroactive |
|
@mliberman See https://github.com/thebrowsercompany/arc/pull/25526 . Swift is inconsistent here >.> |
| @@ -48,7 +48,7 @@ namespace swiftwinrt | |||
| } | |||
| w.write("}\n"); | |||
|
|
|||
| w.write("extension %: ^@retroactive Hashable, ^@retroactive Codable {}\n\n", get_full_swift_type_name(w, type)); | |||
| w.write("extension %: ^@retroactive Hashable, ^@retroactive Codable, ^@retroactive Sendable {}\n\n", get_full_swift_type_name(w, type)); | |||
There was a problem hiding this comment.
@tristanlabelle i.e. I think this needs to be
| w.write("extension %: ^@retroactive Hashable, ^@retroactive Codable, ^@retroactive Sendable {}\n\n", get_full_swift_type_name(w, type)); | |
| w.write("extension %: ^@retroactive Hashable, ^@retroactive Codable, ^@retroactive @unchecked Sendable {}\n\n", get_full_swift_type_name(w, type)); | |
| Comment |
There was a problem hiding this comment.
Yes probably, will update. It's a bummer that Swift is inconsistent.
There was a problem hiding this comment.
🤔 it doesn't seem inconsistent to me, I think the error message is just poorly worded.
There was a problem hiding this comment.
Hmm I think I get it but yeah the message is misleading because @retroactive allows the conformance to be in a different module, which seems like a superset of @unchecked which is described as allowing the conformance to be in a different file.
conformance to 'Sendable' must occur in the same source file as struct '__x_ABI_CArcCoreWinRT_CAdhocHistogramType'; use '@unchecked Sendable' for retroactive conformance; this is an error in the Swift 6 language mode",
WinRT enums are integers in disguise and projected as typealiases for C enums, so they are always Sendable.