Discussion:
C++ problems with clang-cl in math.h
Add Reply
Gisle Vanem
2017-11-02 14:10:41 UTC
Reply
Permalink
Raw Message
Hello list.

I'm trying to use Gnulib's <math.h> from clang-cl C++
compiler. I have no problems with pure C with either clang-cl
or MSVC. But in C++ I get endless troubles with errors like these:

f:/MingW32/src/gnu/gnulib/lib\math.h(2397,36): error: functions that
differ only in their return type cannot be overloaded
_GL_MATH_CXX_REAL_FLOATING_DECL_2 (signbit)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~^~~~~~~~

Thanks to the excellent 'clang-format' program, I was able to generate
this preprocessed output:

// from MSVC and UCRT's <corecrt_math.h>:

inline bool signbit(float _X) throw()
{
return _fdsign(_X) != 0;
}
inline bool signbit(double _X) throw()
{
return _dsign(_X) != 0;
}
inline bool signbit(long double _X) throw()
{
return _ldsign(_X) != 0;
}

// and then from Gnulib's <math.h>:

inline int signbit(float f)
{
return _gl_cxx_signbitf(f);
}
inline int signbit(double d)
{
return _gl_cxx_signbitd(d);
}
inline int signbit(long double l)
{
return _gl_cxx_signbitl(l);
}

I'm getting these error even without a 'REPLACE_SIGNBIT=1'.
And I fail to see how to prevent the above from getting into
<math.h>.

I'm not C++ export, but would imagine that 'int' and 'bool'
are different types. Or does the 'throw()' have anything to do with
this error? Any hints please?

PS. I've tried clang-cl v3.9.0 and 5.0.0 with the same result.
--
--gv
Loading...