We would previously hide the protocol, its extensions and members, but the '_'
prefix really just means the protocol itself isn't intended for clients, rather
than its members.
This also adds support for 'fully_annotated_decl' entries in doc-info for
extensions to be consistent with every other decl, and removes the
'fully_annotated_generic_signature' entry we supplied as a fallback.
Also fixes several bugs with the synthesized extensions mechanism:
- The type sustitutions applied to the extension's requirements were computed
using the extension itself as the decl context rather than the extension's
nominal. The meant the extension's requirements themselves were assumed to
hold when determining the substitutions, so equality constraints were always
met. Because of this extension members were incorrectly merged with the base
nominal or its extensions despite having additional constraints.
- Types within the requirements weren't being transformed when printed (e.g.
'Self.Element' was printed rather than 'T') both in the interface output and
in the requirements list. We were also incorrectly printing requirements
that were already satisfied once the base type was subsituted in.
- If both the protocol extension and 'enabling' extension of the base nominal
that added the protocol conformance had conditional requirements, we were
only printing the protocol extension's requirements in the synthesized
extension.
- The USR and annotated decl output embedded in the 'key.doc.full_as_xml'
string for synthesized members were printed to match their original context, rather than
the synthesized one.
Resolves rdar://problem/57121937
Rather than replacing the code completion file
on the `CompilerInstance` whenever we do a cached
top-level completion, let's set a new main module
instead.
This allows us to properly update the
`LoadedModules` map, and allows the retrieval of
the code completion file to be turned into a
request.
Out handling of clang submodules was handled differently between DocInfo and
InterfaceGen. For InterfaceGen submodules were mapped back to their top-level
clang modules (or their Swift overlay if it had one) before being passed
into printSubmoduleInterface, along with the dot separated name of the submodule.
For DocInfo, they were not, and only the rightmost component of their name was
passed. The call to retrieve the decls from a ModuleDecl doesn't work if the
ModuleDecl wraps a clang submodule, so we were missing these decls.
InterfaceGen for submodules also shouldn't have been mapping the module back to
the overlay of top-level clang module, as that meant we ended up printing
import decls from the Swift overlay in the submodule's interface.
Resolves rdar://problem/57338105
When completing a single argument for a trailing closure, pre-expand the
closure expression syntax instead of using a placeholder. It's not valid
to pass a non-closure anyway.
rdar://62189182
func foo() {}
let a: Int = #^HERE^#
Previously, we marked 'foo()' as 'NotRecommented' because 'Void' doesn't
have any member hence it cannot be 'Int'. But it wass confusing with
'deprecated'.
Now that we output 'typerelation' which is 'invalid' in this case. So clients
can deprioritize results, or even filter them out.
rdar://problem/57726512
Check if dependencies are modified since the last checking.
Dependencies:
- Other source files in the current module
- Dependent files collected by the dependency tracker
When:
- If the last dependency check was over N (defaults to 5) seconds ago
Invalidate if:
- The dependency file is missing
- The modification time of the dependecy is greater than the last check
- If the modification time is zero, compare the content using the file
system from the previous completion and the current completion
rdar://problem/62336432
This change makes us treat it exactly as we do 'init'. We don't allow renaming the base name,
and don't fail if the basename doesn't match for calls.
Also:
- explicit init calls/references like `MyType.init(42)` are now reported with
'init' as a keywordBase range, rather than nothing.
- cursor info no longer reports rename as available on init/callAsFunction
calls without arguments, as there's nothing to rename in that case.
- Improved detection of when a referenced function is a call (rather than
reference) across syntactic rename, cursor-info, and indexing.
Resolves rdar://problem/60340429
A request is intended to be a pure function of its inputs. That function could, in theory, fail. In practice, there were basically no requests taking advantage of this ability - the few that were using it to explicitly detect cycles can just return reasonable defaults instead of forwarding the error on up the stack.
This is because cycles are checked by *the Evaluator*, and are unwound by the Evaluator.
Therefore, restore the idea that the evaluate functions are themselves pure, but keep the idea that *evaluation* of those requests may fail. This model enables the best of both worlds: we not only keep the evaluator flexible enough to handle future use cases like cancellation and diagnostic invalidation, but also request-based dependencies using the values computed at the evaluation points. These aforementioned use cases would use the llvm::Expected interface and the regular evaluation-point interface respectively.
* Reference is marked "explicit", which may be unexpected - the reason
is that the *call* is explicit, so we want to find it with e.g. rename,
or looking up callers, even though the identifier callAsFunction is
implicit. This matches the behaviour of initializers.
* The source location is the same as the base name (e.g. in `add3(5)`,
it would be at `add3`), which matches the behaviour of initializers.
rdar://problem/60327632
* Update DAK_Available name to match @available keyword
Update expected errors for @available attribute tests
* Update available attribute keyword name in DeclAttrKeyword documentation comment
Now
* NotApplicable: The result is not relevant for type relation (e.g.
keywords, and overloads)
* Unknown: the relation was not calculated (e.g. cached results), or the
context type is unknown.
* Invalid: The result type is invalid for this context (i.e. 'Void' for
non-'Void' context)
* Unrelated: The result type has no relation to the context type
* Convertible: The result type is convertible to the context type
* Identical: The result type is identical to the context type
Because we won’t be serializing this attribute, add custom diagnostics for the cases where:
- We add @_hasMissingDesignatedInits to an open class, which means subclasses won’t be able to inherit its inits
- We remove @_inheritsConvenienceInitializers, which means APIs are removed
