ID Number: Q27835
5.00 5.10 5.10a | 5.10 5.10a
MS-DOS | OS/2
buglist5.00 buglist5.10 buglist5.10a fixlist6.00
Summary:
SYMPTOMS
When using scaled indexed addressing in the following Microsoft
Macro Assembler (MASM) 5.0, 5.1, or 5.1a instruction, the opcode
generated is incorrect. Note that the variable "esi" is being used
as an index into a structure and "eax" is the base.
mov eax,[eax.esi*2]
The opcodes generated by MASM indicate that MASM is not using
scaled indexed addressing. The code generated by MASM is
effectively the same as:
mov eax,[eax+esi]
CAUSE
The problem occurs because the "." operator has a higher
precedence than the "*" operator. MASM turns:
[eax.esi*2]
into
[(eax+esi)*2]
as a result of the precedence. MASM evaluates "(eax+esi)" as a
constant instead of a register. Thus, the result of the constant
expression "(eax+esi)" is multiplied by two and the addressing
mode is lost. This results in MASM interpretting the operand as
[eax+esi]
RESOLUTION
One workaround is to use the "+" operator which has lower
precedence than the "." operator and produces the correct result.
Alternately parenthesis can be used to give higher precedence to
the "*" operator.
STATUS
Microsoft has confirmed this to be a problem in MASM version 5.0,
5.1, and 5.1a. This problem was corrected in MASM version 6.0.
More Information:
Under Microsoft MASM 6.0 the following error will be generated.
error A2026: constant expected
Sample Code
-----------
; Assemble options needed: none
.386
_TEXT SEGMENT PARA PUBLIC USE16 'CODE'
ASSUME cs:_TEXT
start:
mov eax,[eax+esi*2] ; This is the desired instruction
; Note the use of "+" instead of "."
mov eax,[eax.(esi*2)] ; This is also the desired instruction
; Note the use of () around the "*"
mov eax,[eax.esi*2] ; This is the format that is misinterpreted
; Under MASM 6.0 this generates an A2026
mov eax,[eax+esi] ; This is what is incorrectly generated
; by MASM 5.0, 5.1, and 5.1a
mov ax, 4C00h
int 21h
_TEXT ENDS
END start