We were implicitly assuming that a function reference could only happen
in an expression, ignoring the case of
import func Module.fooFunc
For now, this doesn't actually add the reference to the index because
initCallRefIndexSymbol doesn't allow references without a parent
expression. We can look at adding the reference, or maybe doing
something special to the import itself separately.
rdar://problem/26496135
A compile-time conditional had previously been introduced to prevent
`private` test methods from being considered test candidates. This
conditional was not working as intended, however, because the header
providing the required preprocessor flag was not being included.
Several number of issues with the non-Objc-interop test case have also
been fixed.
Since the memory underlying returned USRs is guaranteed to last until
the indexing action completes, add a finish() hook so that clients can
trigger an action at the end of indexing but before the memory goes
away.
The goal for https://bugs.swift.org/browse/SR-710 is to automatically
generate a list of tests to run for XCTest on Linux. The prevailing
approach is to generate this list using SourceKit, which is to be
ported to Linux.
SourceKit uses libIndex's concept of `isTestCandidate` to determine
what constitutes a test. `isTestCandidate` is tested via
`test/SourceKit/Indexing/index.swift`.
On Linux, however, the list of tests to be run will be generated by
some tool and placed in a file that is separate from the source file
that defines the test method. Therefore, the test method must not be
"private", since it needs to be accessed from a separate file.
This commit adds two test files: one that verifies the behavior on
Linux, and one that verifies the behavior on platforms with Objective-C
interop. It also (1) simplifies the `isTestCandidate` function, and (2)
adds the interop-specific logic.
This commit does not remove the existing `isTestCandidate` tests in
`test/SourceKit/Indexing/index.swift`; that is left for a future commit.
Fold UnitTest into the subkind, now that it's a bitself, and then remove
the unnecessary inheritance for IndexSymbol and its SourceKit indexing
equivalent.
Eventually we should just have one IndexSymbol, with all the
information. Once we kill FuncDeclIndexSymbol and can get rid of the
enum, we will regain most of the bytes we lost by inlining receiverUSR.
As a first step toward having more detailed "role" information, thread
through a SymbolRole bitset. For now it just contains the existing ref
vs. definition status.
Having a single interface for source files and modules and having to
pass in a buffer id was a crappy API. This splits the interface to take
either a Module or a SourceFile and handles the buffer id stuff
internally.
