Re: unix to dos eol conv.

naunetr wrote:
hi all,

i did this to convert LF sequence to CRLF from stdin to stdout. it appears to work okay. when cpu execute neg eax instruction is code below and eax is 0, will it remain unchanged? or it becomes -0? anything else i do in error?

Looks good to me, and appears to work. Great! I'm getting a warning from Nasm about the "resb" in section .data. Resb, resw, resd, resq, rest, and reso (I think that's all of 'em) should be used in an uninitialized section - .bss or a "custom" section with "nobits" set. In an initialized section "db 0" would be correct (or "times N db 0" for a bigger buffer). Nasm just bitches and zeros it, so it doesn't do any harm. Don't try to initialize data in .bss - Nasm will ignore it and you *won't* get what you want.

; lf2crlf.asm - convert LF to CRLF; from stdin to stdout
section .text

global _start

_start:
mov ecx, buf ; address of read/write buffer
mov edx, 0x01 ; bytes to read/write

Okay, by doing this prior to the call, you can reuse "do_read" to read a different number of bytes into a different buffer, if you want to. Good. (but then your error-checking locks you into one byte...)

call do_read
cmp eax, 0x0 ; success?
jne short do_exit ; failed
call do_write
cmp eax, 0x0
je short _start ; loop again
do_exit:
mov eax, 0x01 ; syscall no. for exit
mov ebx, esi ; load status code from esi
int 0x80
do_read:
mov eax, 0x03 ; syscall no. for read
xor ebx, ebx ; read from fd 0 = stdin
int 0x80

Reusable, so far.

cmp eax, 0x01 ; success?

This locks you into using this routine for just one byte. Lemmee see... "cmp eax, edx" is no good... not an error to read less than we asked for. How about leaving "cmp eax, 1"...

jne short read_fail

.... and making this "jl ..." (signed). That way, 0 (EOF) or an error (negative value) would jump, and "any number" would return success...

xor eax, eax ; ret. zero for sucess
ret

Okay...

read_fail:
neg eax ; 0 is unchanged(?) and -1 becomes 1
mov esi, eax

Okay... you want to return the positive error number - from the program, for "echo $?", not just from this routine - right? The error numbers are defined as positive integers, but what we see in eax (if there's an error) is the negative of that number. But if eax was zero, we just want to return some non-zero value from this routine... but zero from the program. Right?

inc eax ; return failure code
ret

Neat!

do_write:
mov eax, 0x04 ; syscall no. for write
inc ebx ; write to fd 1 = stdout
cmp byte [buf], 0x0a ; is byte LF?
jne short call_write ; no, so output as is
mov ecx, crlf ; yes, so output CRLF sequence
inc edx ; write 2 bytes
int 0x80
cmp eax, 0x02 ; success?
jne short write_fail
jmp short return
call_write:
int 0x80
cmp eax, 0x01 ; okay?
jne short write_fail
return:
xor eax, eax ; success. so return 0
ret
write_fail:
mov esi, 0x01 ; failed. so return 1
mov eax, esi ; and also in eax
ret

section .data

buf resb 0x01
crlf db 0x0d,0x0a

Pretty good! I have a tough time finding anything to complain about.

Best,
Frank
.

Relevant Pages

  • Re: Macro2D
    ... õ1, eax", 0 ... db "push striIIi", NL ... db ".2: cmp edi, ebp", NL ... db ".4: mov eax, edi", NL ...
    (alt.lang.asm)
  • One RosAsm Pre-parser
    ... cmp eax 0 | ja L0<< ... mov ecx, D ...
    (alt.lang.asm)
  • Re: Fastcode Library Design
    ... cmp ecx, SMALLMOVESIZE ... lea edx, ... fild qword ptr [eax] ... mov ecx, ...
    (borland.public.delphi.language.basm)
  • Re: Macro2D
    ... mov, eax ... cmp eax, -1 ... push STD_INPUT_HANDLE ...
    (alt.lang.asm)
  • a problem in n5.dll in windows 95
    ... eax srandMl ... mov ebp, esp ... push dword @lpm ... mov esi, @adw ...
    (alt.lang.asm)