[CodeCompletion] Fix issue causing the completion status to not be set correctly for pattern completion

[ahoppen]

We weren’t setting the code completion token status correctly when parsing patterns with code completion tokens.

Because of this, in the added test case, the searchSubject gets added as a member of Foo twice - once in the first pass and once in the second pass, causing an assertion failure.

Fixes rdar://80575116 [SR-14687]

[ahoppen]

@swift-ci Please smoke test

[hamishknight]

A little unfortunate that this is needed, but looks good to me!

[rintaro]

Hmm, there's a guard not to call Handler in code completion first pass:
https://github.com/apple/swift/blob/35aa862e1d9ab38b530faefdc6a8dd21839b96cf/lib/Parse/ParseDecl.cpp#L4287-L4289

https://github.com/apple/swift/blob/35aa862e1d9ab38b530faefdc6a8dd21839b96cf/lib/Parse/ParseDecl.cpp#L4564-L4565

So if

once in the first pass and once in the second pass.

is true, there must be something wrong in DeclResult status.

[rintaro]

class Foo {
   let searchSubject = Bar<String, #^COMPLETE^#
 }

Apparently, <String, #^COMPLETE^# is not parsed as a generic argument list because it's missing >. So I wonder why the let decl ends up being a completion declaration.. 🤔

[rintaro]

An interesting test case is

protocol P {
  var value: Int { get }
}
struct S: P {
  let foo = val, #COMPLETE^#
}

We want to offer value: Int as an override completion. Currently it crashes in the same way.

[ahoppen]

Thanks a lot, @rintaro, there was indeed a simpler fix to just set the code completion status correctly in pattern parsing. I updated the PR accordingly.

[ahoppen]

@swift-ci Please smoke test

[ahoppen]

@swift-ci Please smoke test

[ahoppen]

@swift-ci Please smoke test macOS

[ahoppen]

@swift-ci Please smoke test macOS