When prefix mapping paths that are used in clang, ensure we are
consistently using the same prefix mapper from clang. This prevents
mismatches that could cause modules to fail to load.
rdar://123324072
This fixes a compiler error that occurred while building `CxxStdlib.swiftmodule` and other projects that use C++ interop: `error: cannot find type 'std' in scope`.
The compiler was creating two distinct `swift::ModuleDecl` instances for the CxxStdlib module while building the CxxStdlib overlay: one as a main module (the module that is currently being compiled), one as a dependency of the main module. This confused the Swift name lookup logic down the road.
This only happened with recent libc++.
See the inline comment for more details.
rdar://122815523
Fix the bridging header dependencies calculation for explicit module
build, especially for caching which needs an accurate list of deps for
compute cache key correctly.
Previously, the bridging header deps are computed from `ModuleGraph`
from the clang dependency scanner, which can be affected by already seen
modules. It causes the dependencies to be missing for bridging header if
the module is seen by main swift source module.
Now report only the directly module dependencies from bridging header,
then compute all the transitive dependencies before calculating all the
cache keys.
rdar://123156636
An @_objcImpl extension with no category name *should* implement not only the class’s main @interface, but also any class extension @interfaces. Start making this true by making ObjCInterfaceAndImplementationRequest return all of these decls as the interfaces for such an implementation.
This commit doesn’t actually change Sema or IRGen to process the extra interfaces, so it’s NFC.
This reduces the memory overhead of objcImpl from one word per Decl to one bit per Decl, at the cost of making cache lookups slightly slower (but it will only be consulted once for non-objcImpl decls, which is by far the most common case).
These constants could be used with observer APIs from a different isolation
context, so it's more convenient to import them as `nonisolated` unless
they are explicitly isolated to a MainActor.
Resolves: rdar://114052705
Previously, the lazily-created extensions used for globals imported as members of a type used the lazy module loader, but `ClangImporter::Implementation::loadNamedMembers()` didn’t actually work for them. Other parts of the compiler instead contrived to avoid loading these members by name by forcing all members to load before any selective loading might occur.
This commit modifies that code path to accommodate extensions with no matching clang node, which is how these are represented. With this change, other parts of the compiler can unconditionally use the `LazyMemberLoader` whenever it is present.
There may be performance improvements from this change, but I don’t expect any functional changes.
