Re: left and right bit shift assembly operators

Indira wrote:

> Since no-one answered my previous post

hey give us some time :)

> I want to add more details in regards to this cirtical problem that 
> we're experiencing. Here's the code snippet (Darwin for i86) that 
> generates a problem (the last line is the one that the compiler throws 
> an error on stating that "Relocation error. Absolute 0 assumed":

> .file "devio.w"
> .text
> .long c.false, c.procedure_key
> .L4:
> .long x.L4, c.false
> .long .L9
> .byte 0, 1
> .byte 0, 0
> .value 0
> .byte 14, 13, 0, 12, 0, 0
> .long .L5
> x.L4:
> movl $.L4, %ebp
> subl $48, %esp
> pushl %ebp
> .L6:
> movl (%ebx), %eax
> movl %eax, 4(%esp)
> leal 8(%esp), %eax
> movl %eax, -4(%ebx)
> movl $2, -8(%ebx)
> movl $tcgetattr, -12(%ebx)
> leal -12(%ebx), %ebx
> call c.031call_sys
> cmpl $0, (%ebx)
> leal 4(%ebx), %ebx
> jns .L8
> movl $c.false, -4(%ebx)
> leal -4(%ebx), %ebx
> jmp .L7
> .L8:
> movl $c.true, -4(%ebx)
> leal -4(%ebx), %ebx
> .L7:
> .L5:
> addl $52, %esp
> movl 4(%esp), %ebp
> ret
> .align 4
> .LA:
> .set .L9, ((.LA - x.L4) >> 2) + 9

> Here is what we know about the problem.
>
> 1) The bitwise right shift operator is the one that it has a problem 
> with. If we cut this operation on 3 different lines, the line that 
> includes the right shift id the one that it will have a problem on.

ok, I cannot claim to know what the problem exactly is ( I hate x86 
assembler ) but I can tell you what bugs me

you are setting the label .L9 to be the address of label .LA ( right 
before this directive ) minus the address of label x.L4 very near the 
beginning of your .text section, you are then shifting the sum right by 
2 ( dividing by 4 ) and adding 9.

Chances are that value is before the .text section, could even be zero.

Because x86 has variable sized instructions, it is pure guess work.

> 2) The bitwise right shift is documented to work on Integers only.

depending of your definition of integer :)

> The defintion of L9 variable is stated to be of type long.

where is that defined ?

> We tried to make this an Integer, but we still experienced the same 
> trouble afterwards.

because it looks like you are storing addresses there

> I hope that you can now tell me something that I can do to fix this 
> problem.

you have to work out what the intention of the original code was

> Original post:
>
> We're porting an i86 software (poplog) from Linux for i86 to Darwin. 
> We're almost done with it, but experience a problem.
>
> Here's the problem: When we have a set instruction that reads as 
> following .set .L9, ((.LA-X.L4)>>2)+9, the right shift operator 
> generates a compile time error stating that " Relocation error. 
> Absolute 0 assumed".
>
> The problem is well identified, but we are somewhat clueless in 
> regards to the solution. Any help is well appreciated.

hope this helps, I could be completely wrong

I am really more familiar with PPC


S a m

-------------------------------------------------------

Drawing on my fine command of language, I said nothing.