AST/IRGen: Accept @_weakLinked on import decls to force weak linkage of symbols from a module
#60414
Conversation
tshortli
force-pushed
the
weak-linked-import
branch
2 times, most recently
from
fdd2d96 to
64181cb
Compare
tshortli
changed the title
@_weakLinked on import decls to force weak linkage of s ymbols from a module@_weakLinked on import decls to force weak linkage of symbols from a module
tshortli
force-pushed
the
weak-linked-import
branch
from
64181cb to
e755814
Compare
tshortli
changed the title
@_weakLinked on import decls to force weak linkage of symbols from a module@_weakLinked on import decls to force weak linkage of symbols from a module
tshortli
force-pushed
the
weak-linked-import
branch
2 times, most recently
from
b963e14 to
4389458
Compare
|
@swift-ci please test |
|
@swift-ci please smoke test |
| bool ModuleDecl::isImportedAsWeakLinked(const Decl *targetDecl) const { | ||
| const auto *declaringModule = targetDecl->getModuleContext(); | ||
| for (auto file : getFiles()) { | ||
| if (file->importsModuleAsWeakLinked(declaringModule)) |
There was a problem hiding this comment.
Should we diagnose some source files use @ _weakLinked import Foo but some others use import Foo?
There was a problem hiding this comment.
Yes, I'm planning to do that in an upcoming PR.
| @@ -0,0 +1,159 @@ | |||
| // RUN: %empty-directory(%t) | |||
There was a problem hiding this comment.
Could you also add a test for adding @_weakLinked import Foo where Foo is a transitive dependency? APIs from Foo are not directly used by the source file, but a direct dependency Bar of the source file has some APIs that trigger inlining Foo's APIs. Those APIs should be weakly linked.
There was a problem hiding this comment.
Good idea, I'll add that test.
There was a problem hiding this comment.
This was a good catch. It turns out that the logic in this PR does not properly change the linkage of symbol references inlined from other modules. I think it's going to require some deeper investigation so I'd like to tackle it in another PR.
There was a problem hiding this comment.
I added a short test case that documents that this currently doesn't work.
…ymbols from a module. The effect of declaring an import `@_weakLinked` is to treat every declaration from the module as if it were declared with `@_weakLinked`. This is useful in environments where entire modules may not be present at runtime. Although it is already possible to instruct the linker to weakly link an entire dylib, a Swift attribute provides a way to declare intent in source code and also opens the door to diagnostics and other compiler behaviors that depend on knowing that all the module's symbols will be weakly linked. rdar://96098097
…ported modules. rdar://96098097
…nsupported on Windows.
…sses @_weakLinked import.
tshortli
force-pushed
the
weak-linked-import
branch
from
5a64127 to
1778a76
Compare
|
@swift-ci please test |
The effect of declaring an import
@_weakLinkedis to treat every declaration from the module as if it were declared with@_weakLinked. This is useful in environments where entire modules may not be present at runtime. Although it is already possible to instruct the linker to weakly link an entire dylib, a Swift attribute provides a way to declare intent in source code and also opens the door to diagnostics and other compiler behaviors that depend on knowing that all the module's symbols will be weakly linked.The
@_weakLinkedattribute on imports is not transitive; if module A imports module B which@_weakLinkedimports module C, it does not follow that A should weakly link symbols from C. Just because B needs to operate on runtimes where C is absent does not mean A needs to also operate on runtimes where C is absent. Therefore the@_weakLinkedattribute is not serialized in binary modules nor is it printed in swiftinterface files.Resolves rdar://96098097.