[LLD] [COFF] Add a separate option for allowing duplicate weak symbols #8944
Conversation
llvm#68077) The MinGW mode (enabled with the flag -lldmingw) does allow duplicate weak symbols. A test in compiler-rt/test/profile/Windows/coverage-weak-lld.cpp does currently enable the -lldmingw flag in an MSVC context, in order to deal with duplicate weak symbols. Add a new, separate, lld specific flag for enabling this. In MinGW mode, this is enabled by default, otherwise it is disabled. This allows making the MinGW mode more restrictive in adding libpaths from the surrounding environment; in MinGW mode, all libpaths are passed explicitly by the compiler driver to the linker, which is attempted in https://reviews.llvm.org/D144084.
|
@swift-ci please test |
|
cc @compnerd, would you mind reviewing, or would you know who else to tag? |
There was a problem hiding this comment.
The cherry-pick is fine. However, I am concerned that this is going to break compatibility with link.exe. We default to link.exe as the linker and have tried to maintain compatibility with the platform linker. This would be akin to saying that you cannot use ld64 on Apple targets.
|
I'm looking into whether we can change coverage to avoid relying on duplicate weak symbols, though I'm not sure it will be an entirely straightforward change. For now though, I think using lld-link with |
This cherry-picks llvm#68077 to
stable/20230725. Swift code coverage relies on the linker being able to handle duplicate weak symbols since LLVM emits a weak symbol for the coverage counters for a given function when it encounters a counter increment instruction, and counter increments can be inlined across translation units. Howeverlink.exedoes not support duplicate weak symbols and, in an effort to maintain compatibility, neither doeslld-linkby default. Upstreamlld-linkdoes however have an option to enable this behavior. Cherry-pick that change tostable/20230725so we can start using it in the Swift driver when coverage is enabled.rdar://129337999