Switch to calling `swift::getAvailabilityConstraintsForDecl()` to get the
unsatisfied availability constraints that should be diagnosed.
This was intended to be NFC, but it turns out it fixed a bug in the recently
introduced objc_implementation_direct_to_storage.swift test. In the test,
the stored properties are as unavailable as the context that is accessing them
so the accesses should not be diagnosed. However, this test demonstrates a
bigger issue with `@objc @implementation`, which is that it allows the
implementations of Obj-C interfaces to be less available than the interface,
which effectively provides an availability checking loophole that can be used
to invoke unavailable code.
Typically, access control denies access to member implementations, so the imported interface decl will be used instead. However, in contexts that permit direct access to stored properties—such as accessors, inits, and deinits—their member implementations are accessible; the compiler then relies on a shadowing rule favoring Swift decls over ObjC decls to eliminate the imported interface decl.
However, there are many rules that are higher-priority than the Swift vs. ObjC decls one. In particular, a recent change to availability checking in #77886 caused a higher-priority rule to begin eliminating member implementations which belonged to unavailable extensions. This caused regressions in projects using `@objc @implementation` with classes that are unavailable in Mac Catalyst.
Introduce a fairly high-priority shadowing rule that favors a member implementation over its interface when both are present (i.e. when direct access to storage is permitted).
Fixes rdar://143582383.
Find all the usages of `--enable-experimental-feature` or
`--enable-upcoming-feature` in the tests and replace some of the
`REQUIRES: asserts` to use `REQUIRES: swift-feature-Foo` instead, which
should correctly apply to depending on the asserts/noasserts mode of the
toolchain for each feature.
Remove some comments that talked about enabling asserts since they don't
apply anymore (but I might had miss some).
All this was done with an automated script, so some formatting weirdness
might happen, but I hope I fixed most of those.
There might be some tests that were `REQUIRES: asserts` that might run
in `noasserts` toolchains now. This will normally be because their
feature went from experimental to upcoming/base and the tests were not
updated.
Today ParenType is used:
1. As the type of ParenExpr
2. As the payload type of an unlabeled single
associated value enum case (and the type of
ParenPattern).
3. As the type for an `(X)` TypeRepr
For 1, this leads to some odd behavior, e.g the
type of `(5.0 * 5).squareRoot()` is `(Double)`. For
2, we should be checking the arity of the enum case
constructor parameters and the presence of
ParenPattern respectively. Eventually we ought to
consider replacing Paren/TuplePattern with a
PatternList node, similar to ArgumentList.
3 is one case where it could be argued that there's
some utility in preserving the sugar of the type
that the user wrote. However it's really not clear
to me that this is particularly desirable since a
bunch of diagnostic logic is already stripping
ParenTypes. In cases where we care about how the
type was written in source, we really ought to be
consulting the TypeRepr.
Automatically detect when shouldMarkAsObjC() will behave differently based on the objcImpl early adopter flag and diagnose it. This works in either direction, although there isn’t anything yet that will emit diag:: objc_implementation_will_become_nonobjc.
NFC except for some minor changes to the wording of notes.
A previous change made it so that an objcImpl extension could declare a `final` or `static` member which overrode an `@objc` superclass method, or witnessed an `@objc` protocol requirement, without itself being `@objc`. Prevent that from happening by allowing processing of `final` members to continue to later codepaths, rather than returning early.
Fixes rdar://136113393.
Before the update to support the new syntax, the decl checker treated private and fileprivate members of objcImpl extensions as non-@objc by default. But SE-0436 specified that private and fileprivate members should be implicitly @objc unless opted out, so when support for the final syntax was added, this behavior was changed retroactively.
Unfortunately, we’ve found that some early adopters depended on the old behavior. Restore some logic deleted by #73309 for early adopter syntax *only*, with a warning telling the developer to put `final` or `@nonobjc` on the declaration if they want to preserve the behavior after updating it.
Also tweaks the ObjCImplementationChecker to ensure this logic will actually be run in time for it to suppress the “upgrade to @objc @implementation” warning, and corrects a couple of regressions arising from that change.
Fixes rdar://135747897.
In #69257, we modified `ObjCReason` to carry a pointer to the @implementation attribute for the `MemberOfObjCImplementationExtension` kind. This made it mark the @implementation attribute as invalid, suppressing diagnostics from the ObjCImplementationChecker.
However, invalidating the attribute *also* causes it to be skipped by serialization. That isn’t a problem if the diagnostics are errors, since we’ll never emit the serialized module, but #74135 softened these diagnostics to warnings for early adopters.
The upshot was that if Swift emitted one of these warnings when it compiled a library, clients of that library would see the objcImpl extension as a normal extension instead. This would cause various kinds of mischief: ambiguous name lookups because implementations weren’t being excluded, overrides failing because an implementation was `public` instead of `open`, asserts and crashes in SILGen and IRGen because stored properties were found in seemingly normal extensions, etc.
Fix this by setting a separate bit on ObjCImplementationAttr, rather than the invalid bit, and modifying the implementation checker to manually suppress many diagnostics when that bit is set.
Fixes rdar://134730183.
In this PR i worked on replacing the word accessor in diagnostics with more user-facing terminologies like setter, getter, didSet Observer, and members based on the context of the message.
in some messages i didn't need to pass DescriptiveDeclKind instead i just changed the text copy itself.
i also updated tests, so you might find it easier to check my changes this way.
Please let me know if there's something i should've done in a better way, and request changes if needed. !
i can squash my commits after reviewing getting the PR Reviewed, just to make it easier to be checked commit by commit
Resolves#55887
`fixDeclarationObjCName()` accepted optional selectors, but crashed if they were None. Make these parameters non-optional, treat an invalid `targetName` selector as “no name”, and adjust callers to pass non-optional values.
Re-enables decl/ext/objc_implementation.swift, which was disabled due to this bug.
Fixes rdar://128683206.
Some editors use diagnostics from SourceKit to replace build issues. This causes issues if the diagnostics from SourceKit are formatted differently than the build issues. Make sure they are rendered the same way, removing most uses of `DiagnosticsEditorMode`.
To do so, always emit the `add stubs for conformance` note (which previously was only emitted in editor mode) and remove all `; add <something>` suffixes from notes that state which requirements are missing.
rdar://129283608
Weaken the minimum deployment target error to a warning for objcImpl adopters using the legacy `@_objcImplementation` syntax until we determine exactly how far back it can be safely used.
Fixes rdar://129225596.
Before the change from @_objcImplementation to @objc @implementation, if a member was unrepresentable in ObjC, it would become implicitly `final`. After that change, this is now an error. We do want a diagnostic here, but we don’t want to break backwards compatibility for early adopters.
Soften the error to a warning when the old @objcImplementation syntax is used.
Fixes rdar://129247349.
…for extensions. This change also removes @implementation(CategoryName); you should attach the category name to the @objc attribute instead. And there are small changes to how much checking the compiler will do on an @objc @implementation after the decl checker has discovered a problem with it.
This now specifies a category name that’s used in TBDGen, IRGen, and PrintAsClang. There are also now category name conflict diagnostics; these subsume some @implementation diagnostics.
(It turns out there was already a check for @objc(CustomName) to make sure it wasn’t a selector!)
This protocol appears in the stdlib as scaffolding for the
`NonescapableTypes` feature, which is still experimental and not gone
through evolution as an approved addition to the stdlib.
Rather than delete it from the stdlib, because it needs to still remain
to support that feature work, gate references to it behind a feature
flag.
Additionally, prevent documentation from seeing this declaration.
rdar://126705184
• ObjCImplementation controls @implementation on extensions
• CImplementation controls @implementation and @_objcImplementation on cdecl functions
Why the difference between them? Because `@_objcImplementation extension` has already been adopted pretty widely, while `@_objcImplementation @_cdecl` is very new.
The recent change to make objcImpl examine class extensions has collided with a workaround some projects were employing where they declared class extension member implementations `private`. Now that we’re checking class extensions in objcImpl extensions, this workaround is creating selector conflict errors in some circumstances. Soften selector conflict errors in objcImpl extensions to warnings so the affected projects continue to build.
Fixes rdar://123347290.
Member implementations are never directly visible to clients of a module, even if they are formally public, but TypeCheckAvailability did not know this. As a result, it would incorrectly diagnose member implementations in an objcImpl extension as violating @_implementationOnly import rules or other, similar access checks. This commit excludes them from being considered as exported declarations.
Fixes rdar://121229058.
ObjCImplementationChecker::matchTypes() implicitly assumed that if it was comparing function types, it must be working on an AbstractFunctionDecl. This isn’t true for a property of block type, which could cause crashes when checking their parameter types. Fix this oversight.
Fixes rdar://122280735.
There's no good reason to permit them. Conformances like Copyable and
Sendable are pervasive, so it's as though we are permitting extensions
of `Any`. Until there's a good argument in favor of such extensions,
remove the capability now.
