bpo-46640: Py_NAN now uses the C99 NAN constant (GH-31134)

Building Python now requires a C99 <math.h> header file providing a
NAN constant, or the __builtin_nan() built-in function. If a platform
does not support Not-a-Number (NaN), the Py_NO_NAN macro can be
defined in the pyconfig.h file.

Doc/whatsnew/3.11.rst

@@ -604,6 +604,12 @@ Build Changes
   ``isinf()``, ``isnan()``, ``round()``.
   (Contributed by Victor Stinner in :issue:`45440`.)
 
+* Building Python now requires a C99 ``<math.h>`` header file providing
+  a ``NAN`` constant, or the ``__builtin_nan()`` built-in function. If a
+  platform does not support Not-a-Number (NaN), the ``Py_NO_NAN`` macro can be
+  defined in the ``pyconfig.h`` file.
+  (Contributed by Victor Stinner in :issue:`46640`.)
+
 * Freelists for object structs can now be disabled. A new :program:`configure`
   option :option:`!--without-freelists` can be used to disable all freelists
   except empty tuple singleton.

Include/pymath.h

@@ -51,29 +51,18 @@
 #endif
 
 /* Py_NAN
- * A value that evaluates to a NaN. On IEEE 754 platforms INF*0 or
- * INF/INF works. Define Py_NO_NAN in pyconfig.h if your platform
- * doesn't support NaNs.
+ * A value that evaluates to a quiet Not-a-Number (NaN).
+ * Define Py_NO_NAN in pyconfig.h if your platform doesn't support NaNs.
  */
 #if !defined(Py_NAN) && !defined(Py_NO_NAN)
-#  if !defined(__INTEL_COMPILER)
-#    define Py_NAN (Py_HUGE_VAL * 0.)
-#  else /* __INTEL_COMPILER */
-#    if defined(ICC_NAN_STRICT)
-        #pragma float_control(push)
-        #pragma float_control(precise, on)
-        #pragma float_control(except,  on)
-        Py_NO_INLINE static double __icc_nan()
-        {
-            return sqrt(-1.0);
-        }
-        #pragma float_control (pop)
-#       define Py_NAN __icc_nan()
-#    else /* ICC_NAN_RELAXED as default for Intel Compiler */
-        static const union { unsigned char buf[8]; double __icc_nan; } __nan_store = {0,0,0,0,0,0,0xf8,0x7f};
-#       define Py_NAN (__nan_store.__icc_nan)
-#    endif /* ICC_NAN_STRICT */
-#  endif /* __INTEL_COMPILER */
+#  if _Py__has_builtin(__builtin_nan)
+     // Built-in implementation of the ISO C99 function nan(): quiet NaN.
+#    define Py_NAN (__builtin_nan(""))
+#else
+     // Use C99 NAN constant: quiet Not-A-Number.
+     // NAN is a float, Py_NAN is a double: cast to double.
+#    define Py_NAN ((double)NAN)
+#  endif
 #endif
 
 #endif /* Py_PYMATH_H */

Misc/NEWS.d/next/Build/2022-02-04-21-26-50.bpo-46640.HXUmQp.rst

@@ -0,0 +1,4 @@
+Building Python now requires a C99 ``<math.h>`` header file providing a ``NAN``
+constant, or the ``__builtin_nan()`` built-in function. If a platform does not
+support Not-a-Number (NaN), the ``Py_NO_NAN`` macro can be defined in the
+``pyconfig.h`` file. Patch by Victor Stinner.