PH63278: OPTIMIZATION GENERATES INCORRECT ASSEMBLER INSTRUCTIONS FOR FUNCTION ARGUMENTS

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• When compiling the following testcase with optimization:
  **>cat structTypes.h**
  ```
  typedef struct twoFloatStruct {
                float elem1;
                        float elem2;
  } twoFloatStruct;
  
  ```
  >
  **>cat test.c**
  ```
  #include <stdio.h>
  #include "structTypes.h"
  
  extern float assignElems(twoFloatStruct arg1, double arg2, int
  arg3);
  
  twoFloatStruct
  test(twoFloatStruct arg1, double arg2, int arg3, void *arg4) {
        twoFloatStruct retStruct;
        retStruct.elem1 = *(float *)arg4;
        retStruct.elem2 = assignElems(arg1, arg2, arg3);
        return retStruct;
  }
  ```
  >
  
  **>cat test2.c**
  ```
  #include <stdio.h>
  #include "structTypes.h"
  
  float
  assignElems(twoFloatStruct arg1, double arg2, int arg3) {
        return arg3 == 5000 ? arg1.elem1 + arg1.elem2 +
  (float)arg2 : -1.0F;
  }
  ```
  
  **>cat run.c**
  ```
  #include <stdio.h>
  #include "structTypes.h"
  
  extern twoFloatStruct test(twoFloatStruct arg1, double arg2, int
  
  arg3, void *arg4);
  
  twoFloatStruct
  testCaller(twoFloatStruct arg1, double arg2, int arg3, void
  *arg4)
  {         return test(arg1, arg2, arg3, arg4);
  }
  
  int main()
  {        twoFloatStruct inputStruct;
        twoFloatStruct retStruct;
        inputStruct.elem1 = 18.23F;
        inputStruct.elem2 = 23.18F;
        float floatInput = 5;
        retStruct = testCaller(inputStruct, 5.65, 5000,
  &floatInput);       printf("Return Elements: elem1: %f, elem2:
  %f, expected:
  5.0F and 47.06F\n", retStruct.elem1, retStruct.elem2);
        return 0;
  }
  ```
  **>**
  
  The assembler code generated from this testcase is incorrect and
  
  does not load arg3 onto the stack. arg1 is a complex-like type
  and therefore passed in FRP0 and FPR2. arg2 is passed in FPR4.
  These both arguments shadow GPR1/2/3, therefore arg3 must be
  passed on the stack.
  

Local fix

• Turn off optimization or do not use 64-bit mode with
  optimization.
  

Problem summary

• ****************************************************************
  * USERS AFFECTED: Users of V2R4 XL C/C++ at 64bit              *
  *                 OPT, passing complex float types (structs    *
  *                 containing                                   *
  *                 two 4 byte floating point type members) as   *
  *                 parameters in a function call.               *
  ****************************************************************
  * PROBLEM DESCRIPTION: If other parameters are passed          *
  *                      in the same function call as the        *
  *                      complex type,                           *
  *                      they may not be passed correctly to     *
  *                      the callee.                             *
  ****************************************************************
  * RECOMMENDATION: Install the PTF                              *
  ****************************************************************
  N/A
  

Problem conclusion

• Install the PTF or use workaround
  

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:

Modules/Macros

• CCNECICS CCNECWI  CCNED240 CCNEDSCT CCNEDWRT CCNEFILT CCNEHIFC
  CCNEIPA3 CCNEIPAO CCNEMDEP CCNEOPTP CCNEP    CCNEPP   CCNETBY
  CCNMSGE  CCNMSGT  CCNQIPA
  

Fix information

• Fixed component name

  C/C++ FOR MVS

• Fixed component ID

  56551210A

Applicable component levels

• R7C0 PSY UI98701

     UP24/10/30 P F410

Fix is available

• Select the PTF appropriate for your component level. You will be required to sign in. Distribution on physical media is not available in all countries.