[SR-7629] Fix wrong fixit when the type doesn't conform to a public protocol #17083
Conversation
|
This looks great! Can you add a test that confirms this works correctly? Off the top of my head, I would do this in test/attr/accessibility.swift:
Like many files in the test suite, attr/accessibility.swift is run in a special mode where the Swift compiler compares the diagnostics emitted by the file to specially-formatted comments. Take a look at other parts of the file to see how it's done. Any diagnostic that isn't mentioned in a comment causes the test to fail, so when you're testing the memberwise behavior, simply not mentioning the fix-it diagnostic in the comments ensures the test will fail if it's emitted. |
|
@brentdax |
|
I wrote tests and checked to pass all tests. |
|
Brent, if you're good with this version so am I; feel free to run tests and merge when you're satisfied with it. (Though I'm sad |
lib/Sema/TypeCheckProtocol.cpp
Outdated
| @@ -2773,6 +2773,12 @@ ConformanceChecker::resolveWitnessViaLookup(ValueDecl *requirement) { | |||
| requiredAccess, | |||
| protoAccessScope.accessLevelForDiagnostics(), | |||
| proto->getName()); | |||
| if (auto *decl = dyn_cast<AbstractFunctionDecl>(witness)) { | |||
| auto isMemberwiseInitializer = decl->getBodyKind() == AbstractFunctionDecl::BodyKind::MemberwiseInitializer; | |||
There was a problem hiding this comment.
Minor nit: can you clang-format this change?
|
@jrose-apple @brentdax Are there other things that you think I should do? |
|
I guess we can take this, and if Brent has further follow-up comments he can still provide them after the merge. |
Wrong fixit occurred when initializer is memberwiseInitializer. So I disabled fixit only in that case.
Resolves SR-7629.