[issue12589] test_long.test_nan_inf() failed on OpenBSD (powerpc)

[STINNER Victor]

STINNER Victor <victor.stinner at haypocalc.com> added the comment:

I'm not sure that your version of gdb understands macros. You may have to set a breakpoint on __isinf. Compile Python with "-g -ggdb" helps gdb.

Py_IS_INFINITY is may not defined as "# define Py_IS_INFINITY(X) isinf(X)". To check that, add #error macro in pymath.h for all cases around #ifndef Py_IS_INFINITY. For example:

#ifndef Py_IS_INFINITY
#  if defined HAVE_DECL_ISINF && HAVE_DECL_ISINF == 1
#    error "Py_IS_INFINITY is implemented using isinf()"
#    define Py_IS_INFINITY(X) isinf(X)
#  else
#    error "Py_IS_INFINITY is implemented using *0.5"
#    define Py_IS_INFINITY(X) ((X) &&                                   \
                               (Py_FORCE_DOUBLE(X)*0.5 == Py_FORCE_DOUBLE(X)))
#  endif
#else
#  error "Py_IS_INFINITY already set!?"
#endif

And recompile Python. It may be a typo in pyconfig.h (generated by configure) or an issue with the inclusion paths of gcc (-I options).

Can you please attach your pyconfig.h and /usr/include/math.h files?

Can you also copy-paste the line used to compile Object/floatobject.c (to check the compiler options)? Use touch Object/floatobject.c && make to recompile this file.

FYI, isinf() is also a macro on Linux:

# ifdef __NO_LONG_DOUBLE_MATH
#  define isinf(x) \
     (sizeof (x) == sizeof (float) ? __isinff (x) : __isinf (x))
# else
#  define isinf(x) \
     (sizeof (x) == sizeof (float)					      \
      ? __isinff (x)							      \
      : sizeof (x) == sizeof (double)					      \
      ? __isinf (x) : __isinfl (x))
# endif

----------

_______________________________________
Python tracker <report at bugs.python.org>
<http://bugs.python.org/issue12589>
_______________________________________