IJ41522: DEFINITION OF INFINITY IN AIX MATH.H DOES NOT ALLOW GLOBAL/STATIC INIT

APAR status

• Closed as program error.

Error description

• When using the AIX math.h header file macro value INFINITY to
  define a variable, error messages are output at compile-time:
  
  C:
  "test.c", line 8.15: 1506-221 (S) Initializer must be a valid
  constant expression.
  
  C++:
  "test.cpp", line 9.28: 1540-0132 (S) The expression must be a
  constant expression.
  "test.cpp", line 9.30: 1540-0063 (S) The text "&" is unexpected.
  
  The variable being defined in C is a global/static value, and
  the variable being defined in C++ is a constexpr value.  In both
  contexts, the INFINITY value needs to be a compile-time
  constant.
  
  
  The definition used by the AIX math.h header file for INFINITY
  is not considered compile-time constant because it generates the
  INFINITY floating point value using a memory
  reference/dereference.
  

Local fix

Problem summary

• PROBLEM DESCRIPTION:
  The macros and definitions in <math.h> for HUGE_VALF, INFINITY
  and NAN have constructs that are not compile-time expressions,
  resulting in errors if they are used in situations where
  compile-time constants are required.
  
  USERS AFFECTED:
  Users of the Legacy xlc/xlC compilers who have global or static
  local variables (C, and if constant initialization is needed by
  the program, C++ as well) or constexpr expressions (C++ with
  -qlanglvl=extended0X) that refer to HUGE_VALF, INFINITY or NAN.
  

Problem conclusion

• New macros are provided for HUGE_VALF, INFINITY and NAN that can
   be used in situations where compile time constants are
  required.  Note that the new definitions are guarded by the
  macro:
  
  __FP_OVERFLOW_BASED_INF_NAN_HUGEVALF
  
  , which must be defined to get the new definitions.  The default
  behaviour (i.e. when this macro is not defined) remains
  unaltered.
  
  It is important to note that the new definitions may cause
  problems by setting floating point error flags.  Also,
  unexpected results may occur in floating point rounding modes
  that round down, notably -ym and -yz.  Should such a rounding
  mode be active at compile time with the macro guard defined when
  <math.h> is processed, the compiler will flag an error on code
  in <math.h>.  The error will indicate a conflict between
  declarations of a variable with a name that describes the
  situation, namely
  __cannot_combine_math_header_config_with_current_rounding_mode__
  .
  

Temporary fix

Comments

APAR Information

• APAR is sysrouted FROM one or more of the following:

• APAR is sysrouted TO one or more of the following:

Fix information

• Fixed component name

  XL C/C++ FOR AI

• Fixed component ID

  5725C7200

Applicable component levels