When 32-bit size value is cast to a far pointer, the upper 16 bits are lost.

This error only occurs when all of the following conditions are fulfilled at the same time.

• 32-bit size constant value is used.
• The above constant is cast to a far pointer.

Only the lower 16 bits of the 32-bit constant are used. The upper 16 bits are ignored.

     ------------------------------------
       char    i;

       void func(void)
       {

              i = *(char far *)0x123456;

       }
     ------------------------------------

     ------------------------------------------------------------------
       ;## # C`SRC :          i = *(char far *)0x123456;
              sem
              lda     A,LG:03456H      /* <-Higher-order is ignored */
              sta     A,DT:_i
     ------------------------------------------------------------------

Please use the #pragma ADDRESS instead of casting the constant value.

Division of double operands, both of which are global variables, may return the wrong result.

This error only occurs when all of the following conditions are fulfilled at the same time.

• Division of two double type variables.
• The variables are both global variables.
• The variables are both declared as near.
• There is an instruction including a sem instruction just before the division.

There is an instruction which is supposed to include clm instruction, but the clm instruction is not there. Therefore the sem condition is held and the operation returns a wrong value. Because clm instruction doesn't exist, the instruction that should be executed after clm is executed by a state of sem.

     ------------------------------------
      double d1, d2;

      double func(char ch)
      {
             double  d = 0.0;

             if( ch == 0 )
                     d = d1 / d2;
             return d;
      }
     ------------------------------------

     ----------------------------------------------------------------------
      ;## # C`SRC :          if( ch == 0 )
             sem
             lda     A,DP:17 ;  ch
             bne     L1
             .cline  8
      ;## # C`SRC :                  d = d1 / d2;

             lda     A,DT:_d1+6   /* <- Data is pushed by a state of sem. */
             pha
             lda     A,DT:_d1+4
             pha
             lda     A,DT:_d1+2
             pha
             lda     A,DT:_d1
             pha
     -----------------------------------------------------------------------

This error only occurs with global variables. Please copy the data into an auto variable and perform the calculations as an auto variable.

When the auto variable is an odd-numbered size, and option -OSFA (-Ostack_fram_align) is specified, the auto variable offset may deviate.

This error only occurs when all of the following conditions are fulfilled at the same time.

• Option -OSFA is specified.
• There is no auto variable which is an odd-numbered size.
• The odd-numbered size auto variable is not used with other auto variables, and therefore, they are overlapped in the stack area.

The auto variable offset value is not correctly set.

     ------------------------------------
      int func(void)
      {
             char c;
             int i;

             i=c;
             return i;
      }
     ------------------------------------

     ------------------------------------------------------------------
      ;## # C`SRC :          i=c;
             sem
             lda     A,DP:0  ;  c        /* <- offset is abnormal */
             clm
             and.W   A,#00ffH
             sta     A,DP:1  ;  i
     ------------------------------------------------------------------

Please do not use option -OSFA.