Difference in stringstream behavior for void* type using libc++ and libstdc++

Question

The following test program returns different results depending on whether I'm using libc++ or libstdc++.

#include <sstream>
#include <iostream>

int main()
{
    int a = 0;
    void* optr = &a;
    void* iptr;

    std::stringstream ss;
    ss << optr;
    std::cout << ss.str() << '\n';

    ss >> iptr;
    std::cout << iptr << '\n';

    return 0;
}

I'm using the following version of clang from Xcode 5 on OSX 10.9.2

$ xcrun clang++ --version              
Apple LLVM version 5.0 (clang-500.2.79) (based on LLVM 3.3svn)
Target: x86_64-apple-darwin13.1.0
Thread model: posix

Here's the output of the test when built with libstdc++ and libc++

$ xcrun clang++ test.cpp  <-- libstdc++ version             
$ ./a.out
0x7fff5ec723e8
0x7fff5ec723e8
$ xcrun clang++ test.cpp -stdlib=libc++  <-- libc++ version
$ ./a.out
0x7fff5205125c
0x7fff5

Is this a bug in the libc++ implementation of stringstream? Is this usage of void* with stringstream valid C++?

Thanks!

Answer 1

Yes! it's a bug in libc++, probably in the implementation of __sscanf_l (some scanf lookalike that should take locales into consideration). The libstdc++ implemementation is much simpler.

// libc++

template <class _CharT, class _InputIterator>
_InputIterator
num_get<_CharT, _InputIterator>::do_get(iter_type __b, iter_type __e,
                                        ios_base& __iob,
                                        ios_base::iostate& __err,
                                        void*& __v) const
{
    // Stage 1
    int __base = 16;
    // Stage 2
    char_type __atoms[26];
    char_type __thousands_sep = 0;
    string __grouping;
    use_facet<ctype<_CharT> >(__iob.getloc()).widen(__num_get_base::__src,
                                                    __num_get_base::__src + 26, __atoms);
    string __buf;
    __buf.resize(__buf.capacity());
    char* __a = &__buf[0];
    char* __a_end = __a;
    unsigned __g[__num_get_base::__num_get_buf_sz];
    unsigned* __g_end = __g;
    unsigned __dc = 0;
    for (; __b != __e; ++__b)
    {
        if (__a_end == __a + __buf.size())
        {
            size_t __tmp = __buf.size();
            __buf.resize(2*__buf.size());
            __buf.resize(__buf.capacity());
            __a = &__buf[0];
            __a_end = __a + __tmp;
        }
        if (this->__stage2_int_loop(*__b, __base, __a, __a_end, __dc,
                                    __thousands_sep, __grouping,
                                    __g, __g_end, __atoms))
            break;
    }
    // Stage 3
    __a[sizeof(__a)-1] = 0;
#ifdef _LIBCPP_LOCALE__L_EXTENSIONS
    if (sscanf_l(__a, _LIBCPP_GET_C_LOCALE, "%p", &__v) != 1)
#else
    if (__sscanf_l(__a, __cloc(), "%p", &__v) != 1)
#endif
        __err = ios_base::failbit;
    // EOF checked
    if (__b == __e)
        __err |= ios_base::eofbit;
    return __b;
}

versus

// libstdc++

template<typename _CharT, typename _InIter>
_InIter
num_get<_CharT, _InIter>::
do_get(iter_type __beg, iter_type __end, ios_base& __io,
       ios_base::iostate& __err, void*& __v) const
{
  // Prepare for hex formatted input.
  typedef ios_base::fmtflags        fmtflags;
  const fmtflags __fmt = __io.flags();
  __io.flags((__fmt & ~ios_base::basefield) | ios_base::hex);

  typedef __gnu_cxx::__conditional_type<(sizeof(void*)
                     <= sizeof(unsigned long)),
unsigned long, unsigned long long>::__type _UIntPtrType;       

  _UIntPtrType __ul;
  __beg = _M_extract_int(__beg, __end, __io, __err, __ul);

  // Reset from hex formatted input.
  __io.flags(__fmt);

  __v = reinterpret_cast<void*>(__ul);
  return __beg;
}

Answer 2

This has been fixed in libc++ as of revision 209305.