backtrace in core on arm-linux

backtrace in core on arm-linux

[Baurzhan Ismagulov]

Hello all,

I can't get the backtrace in a core file on arm-linux. To reproduce:

int main(void) {
	for (;;) printf("Hello, world!\n");
}

Compile, run under gdb HEAD, press Ctrl-\, bt works:

#0  0x400cb530 in write () from /lib/libc.so.6
#1  0x40080c40 in _IO_new_file_write () from /lib/libc.so.6
#2  0x4007f764 in new_do_write () from /lib/libc.so.6
#3  0x4007f86c in _IO_new_do_write () from /lib/libc.so.6
#4  0x400800f0 in _IO_new_file_overflow () from /lib/libc.so.6
#5  0x400816c4 in __overflow () from /lib/libc.so.6
#6  0x40078048 in puts () from /lib/libc.so.6
#7  0x000084bc in main () at b.c:7

Run from shell, press Ctrl-\, open core, bt shows:

#0  0x400cb530 in write () from /lib/libc.so.6
#1  0x00003704 in ?? ()
#2  0x00003704 in ?? ()
Backtrace stopped: previous frame identical to this frame (corrupt stack?)

How can I fix this?

Thanks in advance,
-- 
Baurzhan Ismagulov
http://www.kz-easy.com/

Re: backtrace in core on arm-linux

[Daniel Jacobowitz]

On Mon, Mar 05, 2007 at 07:43:23PM +0100, Baurzhan Ismagulov wrote:
> How can I fix this?

I don't know.  You need to figure out what the difference is - my
first guess would be something wrong with the data your kernel has
saved to the core file (registers or memory).

-- 
Daniel Jacobowitz
CodeSourcery

Re: backtrace in core on arm-linux

[Michael Snyder]

On Mon, 2007-03-05 at 13:52 -0500, Daniel Jacobowitz wrote:
> On Mon, Mar 05, 2007 at 07:43:23PM +0100, Baurzhan Ismagulov wrote:
> > How can I fix this?
> 
> I don't know.  You need to figure out what the difference is - my
> first guess would be something wrong with the data your kernel has
> saved to the core file (registers or memory).

Just as a data point -- a few months ago I tried debugging
an arm-linux core file with gdb, and it seemed to work fine 
then.  I believe I was at gdb-6.5.  I don't specifically
remember doing a backtrace, but it would be surprising if
I didn't do one.

Re: backtrace in core on arm-linux

[Baurzhan Ismagulov]

Hello Daniel and Michael,

Thanks for your fast responses!

On Mon, Mar 05, 2007 at 11:14:26AM -0800, Michael Snyder wrote:
> Just as a data point -- a few months ago I tried debugging
> an arm-linux core file with gdb, and it seemed to work fine 
> then.  I believe I was at gdb-6.5.  I don't specifically
> remember doing a backtrace, but it would be surprising if
> I didn't do one.

That's interesting. I've tried 6.4-debian and 6.6-debian, both didn't
work for me. What was your kernel version? I suppose core file writing
shouldn't be different for different arm machine types, right?

With kind regards,
-- 
Baurzhan Ismagulov
http://www.kz-easy.com/

Re: backtrace in core on arm-linux

[Michael Snyder]

On Tue, 2007-03-06 at 05:56 +0100, Baurzhan Ismagulov wrote:
> Hello Daniel and Michael,
> 
> Thanks for your fast responses!
> 
> On Mon, Mar 05, 2007 at 11:14:26AM -0800, Michael Snyder wrote:
> > Just as a data point -- a few months ago I tried debugging
> > an arm-linux core file with gdb, and it seemed to work fine 
> > then.  I believe I was at gdb-6.5.  I don't specifically
> > remember doing a backtrace, but it would be surprising if
> > I didn't do one.
> 
> That's interesting. I've tried 6.4-debian and 6.6-debian, both didn't
> work for me. What was your kernel version?

2.6.14
>
>  I suppose core file writing
> shouldn't be different for different arm machine types, right?

Don't know, but shouldn't think so.