https://gcc.gnu.org/bugzilla/show_bug.cgi?id=106042
Bug ID: 106042
Summary: std::__once_call abi umatch for clang and gcc on
windows targets
Product: gcc
Version: 13.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: libstdc++
Assignee: unassigned at gcc dot gnu.org
Reporter: unlvsur at live dot com
Target Milestone: ---
Created attachment 53180
--> https://gcc.gnu.org/bugzilla/attachment.cgi?id=53180&action=edit
gcc output
Consider code like this:
extern __thread void* __once_callable;
extern __thread void (*__once_call)() noexcept;
void foo() noexcept
{
__once_call();
}
GCC:
movq .refptr.__emutls_v.__once_call(%rip), %rcx
call __emutls_get_address
movq (%rax), %rax
addq $40, %rsp
rex.W jmp *%rax
but clang --target=x86_64-windows-gnu:
_Z3foov: # @_Z3foov
movl _tls_index(%rip), %eax
movq %gs:88, %rcx
movq (%rcx,%rax,8), %rax
rex64 jmpq *__once_call@SECREL32(%rax) # TAILCALL
This means clang does not have the same behavior for thread_local storage with
GCC on windows.
It breaks clang Canadian compilation from clang with libstdc++. I think we need
some tweaks here.
ld.lld: error: undefined symbol: std::__once_call
>>> referenced by
>>> lib/Support/CMakeFiles/LLVMSupport.dir/Parallel.cpp.obj:(std::__future_base::_State_baseV2::_M_set_result(std::function<std::unique_ptr<std::__future_base::_Result_base,
>>> std::__future_base::_Result_base::_Deleter> ()>, bool))
>>> referenced by
>>> lib/Support/CMakeFiles/LLVMSupport.dir/Parallel.cpp.obj:(std::__future_base::_State_baseV2::_M_set_result(std::function<std::unique_ptr<std::__future_base::_Result_base,
>>> std::__future_base::_Result_base::_Deleter> ()>, bool))
clang-15: error: linker command failed with exit code 1 (use -v to see
invocation)
[61/4971] Building CXX object
lib/FileCheck/CMakeFiles/LLVMFileCheck.dir/FileCheck.cpp.obj
ninja: build stopped: subcommand failed.
