special floats generate incorrect C code when used as compile-time constants

[Jacques Frechet]

For example:

  DEF INFINITY = float('inf')
  print INFINITY

results in:

  foo.c: In function ‘init2foo’:
  foo.c:158: error: ‘inf’ undeclared (first use in this function)
  foo.c:158: error: (Each undeclared identifier is reported only once
  foo.c:158: error: for each function it appears in.)

I get the same error message with -inf and a somewhat different error message with nan (since math.h defines a function by that name).

In the meantime, I'll do this instead:

  cdef extern from "math.h":
    double INFINITY, NAN

[scoder (scoder)]

Fixed in cython-devel. Thanks!

[Christian Heimes (heimes)]

I recommend against using the constants from math.h. Python has some helpers in pyport.h (or pymath.h starting with 2.6):

inf = Py_HUGE_VAL
-inf = -Py_HUGE_VAL
nan = Py_HUGE_VAL * 0.

[scoder (scoder)]

Ah, thanks, I didn't know those constants. Seems to work just fine for me.

http://hg.cython.org/cython-devel/rev/7301088eaaf8
http://hg.cython.org/cython-devel/file/7301088eaaf8/tests/run/specialfloat.pyx