1. Options including optimizing level 2 (-O2, -O3, -O6, -O7, -O, -Otime only, and -Ospase only) are given at compiling.
2. A constant, variable, bit field member, or operation including them is assigned to another bit field member. Also, each of their values, except for the operation of 1's complement, is out of the range of the values the bit field member to which it has been assigned can express. In the operation of 1's complement, the result is any one of the following values:
   • It is out of the range of the values the bit field member to which the assignment has been done can express.
   • Although it is within the above range, the right side term of the expression is unsigned, and references are made to include the outside of the above range.
   • Although it is within the above range, the right side term of the expression is unsigned, and the reference in condition (6) is made to ask whether the value of the bit field member is true or false if it becomes 0.
3. The data type of the bit field member described in (2) is char, unsigned char, short, or unsigned short.
4. Type modifier volatile is not declared to the bit field member described in (2) or its structure.
5. The bit width of the bit field member described in (2) is 15 (1 through 15), except that if the data type of the bit field member is short or unsigned short, the bit width is 8 (8 through 15).
6. The bit field member described in (2) is referenced.
7. No function call, branch statement, or assignment to other bit field member is described between the assignment in (2) and reference in (6).

[Description Example]------------------------------------------------
  struct st { 
	unsigned short	a:   10;  /* <- conditions (3), (4), (5) */
  } S;
  unsigned short  x;
  void foo ( void )
  {
	S.a = 0xff80; /* <- condition (2) */
		      /* <- condition (7) */
  	x = S.a	      /* <- condition (6): x becomes just 0x380 */
  }
---------------------------------------------------------------------

1. The -pack option (structure packing function) is given at compiling.
2. Bit fields are included in members of a structure.
3. The remainder of the sum of the bit lengths specified by contiguous bit field members divided by 32 is between 17 and 24.
4. This structure is statically initialized.

[Description Example]-----------------------------------------------
  struct st {
	unsigned int a:8;  /* a, b, and c are contiguous values.     */
	unsigned int b:8;  /* The sum of the bit lengths is          */
	unsigned int c:8;  /* 8 + 8 + 8 = 24, and the remainder of   */
	                   /* the sum divided by 32 is 24 % 32 = 24. */
  } A = { 1, 1, 1 }
--------------------------------------------------------------------

1. The -RBPP option (using a register to pass an argument to a function) is given at compiling.
2. The type of the argument of a function is char, unsigned char, short, or unsigned short.
3. Before updating the argument described in (2), its address is referenced.
4. Data is referenced using the address in (3)

[Description Example]------------------------------------------------
  #include "stdio.h"
  void test(short *d)
  {
	printf("%x", *d);  /* Condition (4): Correctly referenced */
  }
  void foo(short data)  /* Condition (2) */
  {
	test(&data);  /* Condition (3) */
  }
---------------------------------------------------------------------

[Function of Linker]

When an operand containing the external reference symbol or relative symbol of a displacement not determined prior to linking is used in an assembly language source, the assembler generates an object file containing undetermined areas. On the other hand, the linker relocates determined addresses to these areas to generate the load module with determined addresses.

[Specifications of Overflow Check of Relocation Size]

When relocating a determined address into an undetermined area, the linker examines whether the area has an enough space for relocating the address or not, and if not, it outputs an error message. The previous version assumes the message to be a warning, whereas the upgraded version assumes it to be an error and has, in addition, enlarged the scope of errors (see Tables 1 and 2).