Gcc thread sanitizer false positive only for debug info flag

萝らか妹 提交于 2019-12-22 12:51:03

问题


I am having a problem with Gcc's thread sanitizer that I cannot find on their bugzilla or on stackoverflow so I am unsure if I am missing something or if this really is a bug. If I create a main.cpp file containing:

#include <thread>
int main(){
    std::thread t([](){});
    t.join();
    return 0;}

Now if I compile it using:

g++-4.9.2 -std=c++1y -fsanitize=thread -fPIE -pie -o TestProgram main.cpp

Running the resulting executable does not yield any problem. Yet if I add the debug info flag:

g++-4.9.2 -std=c++1y -fsanitize=thread -g -fPIE -pie -o TestProgram main.cpp

then the thread sanitizer detects a data race:

WARNING: ThreadSanitizer: data race (pid=22683)
  Write of size 8 at 0x7d0c0000efd8 by thread T1:
    #0 operator delete(void*) ../../../../gcc-4.9.2/libsanitizer/tsan/tsan_interceptors.cc:592 (libtsan.so.0+0x000000049490)
    #1 deallocate /usr/local/include/c++/4.9.2/ext/new_allocator.h:110 (TestProgram+0x000000002089)
    #2 deallocate /usr/local/include/c++/4.9.2/bits/alloc_traits.h:383 (TestProgram+0x000000001f78)
    #3 _M_destroy /usr/local/include/c++/4.9.2/bits/shared_ptr_base.h:535 (TestProgram+0x0000000026f4)
    #4 std::_Sp_counted_base<(__gnu_cxx::_Lock_policy)2>::_M_release() /home/UserG/Compile/objdir/x86_64-unknown-linux-gnu/libstdc++-v3/include/bits/shared_ptr_base.h:166 (libstdc++.so.6+0x0000000b5c51)
    #5 ~__shared_count /home/UserG/Compile/objdir/x86_64-unknown-linux-gnu/libstdc++-v3/include/bits/shared_ptr_base.h:666 (libstdc++.so.6+0x0000000b5c51)
    #6 ~__shared_ptr /home/UserG/Compile/objdir/x86_64-unknown-linux-gnu/libstdc++-v3/include/bits/shared_ptr_base.h:914 (libstdc++.so.6+0x0000000b5c51)
    #7 ~shared_ptr /home/UserG/Compile/objdir/x86_64-unknown-linux-gnu/libstdc++-v3/include/bits/shared_ptr.h:93 (libstdc++.so.6+0x0000000b5c51)
    #8 execute_native_thread_routine ../../../../../gcc-4.9.2/libstdc++-v3/src/c++11/thread.cc:95 (libstdc++.so.6+0x0000000b5c51)

  Previous atomic write of size 4 at 0x7d0c0000efd8 by main thread:
    #0 __tsan_atomic32_fetch_add ../../../../gcc-4.9.2/libsanitizer/tsan/tsan_interface_atomic.cc:468 (libtsan.so.0+0x0000000206ce)
    #1 __exchange_and_add /usr/local/include/c++/4.9.2/ext/atomicity.h:49 (TestProgram+0x0000000014a0)
    #2 __exchange_and_add_dispatch /usr/local/include/c++/4.9.2/ext/atomicity.h:82 (TestProgram+0x000000001557)
    #3 std::_Sp_counted_base<(__gnu_cxx::_Lock_policy)2>::_M_release() /usr/local/include/c++/4.9.2/bits/shared_ptr_base.h:146 (TestProgram+0x000000002ceb)
    #4 std::__shared_count<(__gnu_cxx::_Lock_policy)2>::~__shared_count() /usr/local/include/c++/4.9.2/bits/shared_ptr_base.h:666 (TestProgram+0x000000002cb6)
    #5 std::__shared_ptr<std::thread::_Impl_base, (__gnu_cxx::_Lock_policy)2>::~__shared_ptr() /usr/local/include/c++/4.9.2/bits/shared_ptr_base.h:914 (TestProgram+0x000000002bc1)
    #6 std::shared_ptr<std::thread::_Impl_base>::~shared_ptr() /usr/local/include/c++/4.9.2/bits/shared_ptr.h:93 (TestProgram+0x000000002bed)
    #7 thread<main()::<lambda()> > /usr/local/include/c++/4.9.2/thread:135 (TestProgram+0x0000000016f1)
    #8 main /home/UserG/main.cpp:3 (TestProgram+0x0000000015af)

  Location is heap block of size 48 at 0x7d0c0000efd0 allocated by main thread:
    #0 operator new(unsigned long) ../../../../gcc-4.9.2/libsanitizer/tsan/tsan_interceptors.cc:560 (libtsan.so.0+0x0000000496d2)
    #1 allocate /usr/local/include/c++/4.9.2/ext/new_allocator.h:104 (TestProgram+0x000000001fe9)
    #2 allocate /usr/local/include/c++/4.9.2/bits/alloc_traits.h:357 (TestProgram+0x000000001ecd)
    #3 __shared_count<std::thread::_Impl<std::_Bind_simple<main()::<lambda()>()> >, std::allocator<std::thread::_Impl<std::_Bind_simple<main()::<lambda()>()> > >, std::_Bind_simple<main()::<lambda()>()> > /usr/local/include/c++/4.9.2/bits/shared_ptr_base.h:616 (TestProgram+0x000000001d99)
    #4 __shared_ptr<std::allocator<std::thread::_Impl<std::_Bind_simple<main()::<lambda()>()> > >, std::_Bind_simple<main()::<lambda()>()> > /usr/local/include/c++/4.9.2/bits/shared_ptr_base.h:1090 (TestProgram+0x000000001ccb)
    #5 shared_ptr<std::allocator<std::thread::_Impl<std::_Bind_simple<main()::<lambda()>()> > >, std::_Bind_simple<main()::<lambda()>()> > /usr/local/include/c++/4.9.2/bits/shared_ptr.h:316 (TestProgram+0x000000001c5f)
    #6 allocate_shared<std::thread::_Impl<std::_Bind_simple<main()::<lambda()>()> >, std::allocator<std::thread::_Impl<std::_Bind_simple<main()::<lambda()>()> > >, std::_Bind_simple<main()::<lambda()>()> > /usr/local/include/c++/4.9.2/bits/shared_ptr.h:588 (TestProgram+0x000000001bf0)
    #7 make_shared<std::thread::_Impl<std::_Bind_simple<main()::<lambda()>()> >, std::_Bind_simple<main()::<lambda()>()> > /usr/local/include/c++/4.9.2/bits/shared_ptr.h:604 (TestProgram+0x000000001ab0)
    #8 _M_make_routine<std::_Bind_simple<main()::<lambda()>()> > /usr/local/include/c++/4.9.2/thread:193 (TestProgram+0x000000001919)
    #9 thread<main()::<lambda()> > /usr/local/include/c++/4.9.2/thread:135 (TestProgram+0x0000000016bf)
    #10 main /home/UserG/main.cpp:3 (TestProgram+0x0000000015af)

  Thread T1 (tid=22685, running) created by main thread at:
    #0 pthread_create ../../../../gcc-4.9.2/libsanitizer/tsan/tsan_interceptors.cc:877 (libtsan.so.0+0x000000047c03)
    #1 __gthread_create /home/UserG/Compile/objdir/x86_64-unknown-linux-gnu/libstdc++-v3/include/x86_64-unknown-linux-gnu/bits/gthr-default.h:662 (libstdc++.so.6+0x0000000b5d00)
    #2 std::thread::_M_start_thread(std::shared_ptr<std::thread::_Impl_base>) ../../../../../gcc-4.9.2/libstdc++-v3/src/c++11/thread.cc:142 (libstdc++.so.6+0x0000000b5d00)
    #3 main /home/UserG/main.cpp:3 (TestProgram+0x0000000015af)

SUMMARY: ThreadSanitizer: data race /usr/local/include/c++/4.9.2/ext/new_allocator.h:110 deallocate
==================
ThreadSanitizer: reported 1 warnings

Now the exact same code compiled with clang++ (version 3.6.0 (trunk 221144)) does not detect a data race:

clang++ -std=c++1y -fsanitize=thread -g -fPIE -pie -o TestProgram main.cpp

I am a bit quizzical about this behavior from gcc as: 1) passing an empty lambda function as an argument to a thread seems licit to me 2) gcc's behavior depends on the -g flag which doesn't strike me as having much to do with the thread sanitizer 3) under similar circumstances clang adopts a behavior that I would consider correct

Many thanks,


回答1:


You mean this bug? It was reported for 4.8, but a similar report exists for 4.9.2.



来源:https://stackoverflow.com/questions/28719796/gcc-thread-sanitizer-false-positive-only-for-debug-info-flag

易学教程内所有资源均来自网络或用户发布的内容,如有违反法律规定的内容欢迎反馈
该文章没有解决你所遇到的问题?点击提问,说说你的问题,让更多的人一起探讨吧!