MemorySanitizer: use-of-uninitialized-value

[springmeyer]

Seeing this with clang 3.9 on linux and a program built against catch with -fsanitize=memory:

==16495==WARNING: MemorySanitizer: use-of-uninitialized-value
    #0 0x51cd11 in std::_Rb_tree<std::string, std::pair<std::string const, Catch::Ptr<Catch::IReporterFactory> >, std::_Select1st<std::pair<std::string const, Catch::Ptr<Catch::IReporterFactory> > >, std::less<std::string>, std::allocator<std::pair<std::string const, Catch::Ptr<Catch::IReporterFactory> > > >::_M_get_insert_unique_pos(std::string const&) /usr/lib/gcc/x86_64-linux-gnu/4.9/../../../../include/c++/4.9/bits/stl_tree.h:1442:7
    #1 0x51c7f3 in std::pair<std::_Rb_tree_iterator<std::pair<std::string const, Catch::Ptr<Catch::IReporterFactory> > >, bool> std::_Rb_tree<std::string, std::pair<std::string const, Catch::Ptr<Catch::IReporterFactory> >, std::_Select1st<std::pair<std::string const, Catch::Ptr<Catch::IReporterFactory> > >, std::less<std::string>, std::allocator<std::pair<std::string const, Catch::Ptr<Catch::IReporterFactory> > > >::_M_insert_unique<std::pair<std::string, Catch::Ptr<Catch::IReporterFactory> > >(std::pair<std::string, Catch::Ptr<Catch::IReporterFactory> >&&) /usr/lib/gcc/x86_64-linux-gnu/4.9/../../../../include/c++/4.9/bits/stl_tree.h:1498:4
    #2 0x51c4ea in std::pair<std::_Rb_tree_iterator<std::pair<std::string const, Catch::Ptr<Catch::IReporterFactory> > >, bool> std::map<std::string, Catch::Ptr<Catch::IReporterFactory>, std::less<std::string>, std::allocator<std::pair<std::string const, Catch::Ptr<Catch::IReporterFactory> > > >::insert<std::pair<std::string, Catch::Ptr<Catch::IReporterFactory> >, void>(std::pair<std::string, Catch::Ptr<Catch::IReporterFactory> >&&) /usr/lib/gcc/x86_64-linux-gnu/4.9/../../../../include/c++/4.9/bits/stl_map.h:638:23
    #3 0x51c348 in Catch::ReporterRegistry::registerReporter(std::string const&, Catch::Ptr<Catch::IReporterFactory> const&) /home/travis/build/mapbox/wagyu/./tests/catch.hpp:6666:25
    #4 0x4b39cb in Catch::(anonymous namespace)::RegistryHub::registerReporter(std::string const&, Catch::Ptr<Catch::IReporterFactory> const&) /home/travis/build/mapbox/wagyu/./tests/catch.hpp:6771:36
    #5 0x4c6910 in Catch::ReporterRegistrar<Catch::JunitReporter>::ReporterRegistrar(std::string const&) /home/travis/build/mapbox/wagyu/./tests/catch.hpp:8881:37
    #6 0x41e96e in __cxx_global_var_init.67 /home/travis/build/mapbox/wagyu/./tests/catch.hpp:9547:5
    #7 0x41edfc in _GLOBAL__sub_I_test.cpp /home/travis/build/mapbox/wagyu/./tests/test.cpp
    #8 0x763a9c in __libc_csu_init (/home/travis/build/mapbox/wagyu/test+0x763a9c)
    #9 0x7faee215377f  (/lib/x86_64-linux-gnu/libc.so.6+0x2177f)
    #10 0x420a98 in _start (/home/travis/build/mapbox/wagyu/test+0x420a98)

[springmeyer]

If anyone has ideas on how to fix this I would love to hear them. It is very hard to suppress errors from MSan unless you build your own standard library with Msan (https://llvm.org/bugs/show_bug.cgi?id=30740), so fixing this in catch.hpp would be ideal.

[horenmar]

This is most likely a false positive caused by uninstrumented libstdc++/libc++.

If there is something Catch can do to workaround it, we would be willing to implement it, but it is fairly unlikely.

[horenmar]

We could potentially try messing around with __msan_unpoison, but it would be very, very, very low on the priority list. I am going to close this iseue and tag it with revisit.

[hsaturn]

I have the same problem but I do use -stdlib=libc++

    linux-vdso.so.1 (0x00007ffefc74f000)
    libc++.so.1 => /lib/x86_64-linux-gnu/libc++.so.1 (0x00007ff9cc22f000)
    libunwind.so.1 => /lib/x86_64-linux-gnu/libunwind.so.1 (0x00007ff9cc224000)
    libc++abi.so.1 => /lib/x86_64-linux-gnu/libc++abi.so.1 (0x00007ff9cc1ee000)
    libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007ff9cc107000)
    libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007ff9cc0e7000)
    libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007ff9cbebf000)
    /lib64/ld-linux-x86-64.so.2 (0x00007ff9cde54000)

Could anyone explain how to get rid of the error ??? I saw some bug reports, closed, but never how to fix this ?
I think but am not sure that this is related to how libc++ was compiled.... maybe...
It would be kind if instead of just closing or linking to duplicates (that do not show ore explain more) one could take time to simply guide us.

• What options to pass to clang (-fsanitize=memory -fsanitize-recover=memory for me)
• What options to compiler (only -stdlib=libc++ for me)
• What are the prerequisite on libc++ compiling (I'm under ubuntu and used apt install libc++-14-dev libc++abi-dev

I give up after hours, maybe I ws unlucky in my searches but I have to go forward. So now vim will delete for me paragraphs that are poisonning me (g/something/norm dap)

Best regards

[horenmar]

You need to recompile your stdlib with -fsanitize=memory enabled, and then link against that recompiled stdlib.