On Fri, 10 Jul 2015, Andrew Haley wrote:
> On 07/10/2015 06:03 PM, xparmenides wrote:
> > 1. To debug at assembly level, the steps to compile is so complicated
> > that is somewhat impractical. Is there any way make the process easily?
>
> Yes. Compile your program C in the normal way, then:
>
if you compile your c files with
gcc -S --verbose-asm file.c
you will get much more readable assembly output than if you use the
normal disasseble output within gcc.
e.g:
<snip>
devkmsg_poll:
pushq %rbp #
movl $40, %eax #, D.38788
movq %rsp, %rbp #,
pushq %rbx #
subq $8, %rsp #,
movq 208(%rdi), %rbx # file_3(D)->private_data, user
testq %rbx, %rbx # user
je .L27 #,
testq %rsi, %rsi # wait
je .L28 #,
movq (%rsi), %rax # wait_7(D)->_qproc, D.41321
testq %rax, %rax # D.41321
<snip>
telling you what is being put in what register and what intermediate
values gcc added in (like D.41321 here)
furhter there are methods to interleave C and assembly code - atleast in
the linux kernel build system make path/file.lst will generate
the interleaved assembly/C files
e.g.
<snip>
static ssize_t msg_print_ext_body(char *buf, size_t size,
char *dict, size_t dict_len,
char *text, size_t text_len)
{
ffffffff81f40f55: 41 81 fc 00 00 02 00 cmp $0x20000,%r12d
char *p = buf, *e = buf + size;
size_t i;
/* escape non-printable characters */
for (i = 0; i < text_len; i++) {
ffffffff81f40f5c: 76 44 jbe ffffffff81f40fa2 <setup_log_buf+0x94>
ffffffff81f40f5e: be 00 10 00 00 mov $0x1000,%esi
return scnprintf(buf, size, "%u,%llu,%llu,%c;",
(msg->facility << 3) | msg->level, seq, ts_usec, cont);
}
<snip>
see the Linux kernel scripts/makelst for more.
notably when debugging optimized code this output makes it much easier to
read the disassembler output in gdb.
thx!
hofrat
