[PATCH] D96175: [clang] Add support for attribute 'swift_async_error'

[Erik Pilkington via Phabricator via cfe-commits]

erik.pilkington added inline comments.


================
Comment at: clang/lib/Sema/SemaDeclAttr.cpp:5977
+    checkSwiftAsyncErrorBlock(S, D, ErrorAttr, AsyncAttr);
+}
+
----------------
aaron.ballman wrote:
> Should there be a diagnostic to combine `swift_error` and `swift_async_error` on the same function? Similarly, should there be a diagnostic when adding `swift_async_error` to a function that isn't (eventually) marked `swift_async`?
My understanding is that `swift_error` and `swift_async_error` don't conflict with each other. `swift_error` describes how the non-async swift function handles errors, whereas `swift_async_error` describes how the swift async function handles errors. Since both swift functions are generated, I think it can make sense to have both attributes on one ObjC function. 

I think it can make sense to have `swift_async_error` without `swift_async`, since the swift importer infers some functions are async without the `swift_async` attribute using heuristics based on parameter names. I guess we could mimic those heuristics here to diagnose when `swift_async_error` doesn't make sense alone, but ISTM that it makes more sense to do that check in the swift importer.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D96175/new/

https://reviews.llvm.org/D96175