This is the mail archive of the
gcc-help@gcc.gnu.org
mailing list for the GCC project.
Re: tm vs. std:.tm type in symbol name of std::time_put<>::put()
OK, thanks.
Is there a workaround for this issue on the gcc user side? (I guess I could
always avoid using libstdc++ and go with libc++ instead.) I don't have any
constraints when it comes to ABI breakage.
Regards,
/Johan Alfredsson
On Fri, Sep 15, 2017 at 3:34 PM, Marc Glisse <marc.glisse@inria.fr> wrote:
> On Fri, 15 Sep 2017, Marc Glisse wrote:
>
> On Fri, 15 Sep 2017, Johan Alfredsson wrote:
>>
>> I ran into a symbol name difference linking an executable compiled with
>>> clang 5.0 vs. compiling with gcc 7.1.0 on Solaris/SPARC.
>>>
>>> In bits/locale_facets_nonio.tcc, there is the following template:
>>>
>>> template<typename _CharT, typename _OutIter>
>>> _OutIter
>>> time_put<_CharT, _OutIter>::
>>> put(iter_type __s, ios_base& __io, char_type __fill, const tm* __tm,
>>> const _CharT* __beg, const _CharT* __end) const
>>>
>>> Looking at the symbol in libstdc++.so corresponding to the instantiation
>>> of
>>> time_put<char, std::ostreambuf_iterator<char, std::char_traits<char>>>,
>>> it
>>> is
>>>
>>> _ZNKSt8time_putIcSt19ostreambuf_iteratorIcSt11char_
>>> traitsIcEEE3putES3_RSt8ios_basePK2tmPKcSB_
>>>
>>> This corresponds to (according to c++filt):
>>>
>>> std::time_put<char, std::ostreambuf_iterator<char,
>>> std::char_traits<char> >
>>>
>>>> ::put(std::ostreambuf_iterator<char, std::char_traits<char> >,
>>>>
>>> std::ios_base&, tm const*, char const*, char const*) const
>>>
>>> Now, linking the executable compiled with clang 5.0 results in an error
>>> about a missing symbol. That symbol is:
>>>
>>> std::time_put<char, std::ostreambuf_iterator<char,
>>> std::char_traits<char> >
>>>
>>>> ::put(std::ostreambuf_iterator<char, std::char_traits<char> >,
>>>>
>>> std::ios_base&, std::tm const*, char const*, char const*) const
>>>
>>> where the difference is that the type tm is now std::tm.
>>>
>>> It seems to me that std::tm should be the proper type here, but I might
>>> be
>>> wrong.
>>>
>>> Comments?
>>>
>>
>> This was done on purpose when we changed __cplusplus from 1 to 199711, to
>> avoid breaking the ABI. On Solaris, we will mangle std::tm as if it was
>> ::tm until the next ABI break (probably not soon).
>>
>
> https://bugs.llvm.org/show_bug.cgi?id=33767
>
> --
> Marc Glisse
>