Incomplete backtraces

[Mohammed Khoory]

Interesting. I would need to recompile GDB and that might be hard under
MinGW though. I'll give it a try nevertheless, thanks.

I've been looking at the SPARC GDB bug tracker and I found some bugs similar
to this one, but they're several years old (2004 or something) and haven't
been fixed yet.

I guess not a lot of people use GDB with SPARC.

>-----Original Message-----
>From: Jiri Gaisler [mailto:jiri at gaisler.se]
>Sent: Wednesday, November 20, 2013 7:30 AM
>To: Mohammed Khoory; rtems-users at rtems.org
>Subject: Re: Incomplete backtraces
>
>On 11/15/2013 05:56 AM, Mohammed Khoory wrote:
>> Hi,
>>
>> I'm not sure if this is really RTEMS-related, but I've been having an
>> annoying issue when debugging an RTEMS application with GDB (on real
>> hardware). Sometimes, when I pause the debugger, or when there is a
>> segfault or something similar, I notice that the stack trace is
>> missing a few function calls deep down. I know this is not a stack
>> issue, because I am able to step out of a function and the system runs
>> stable otherwise. When I step out of a function, the backtrace
>> sometimes becomes complete again. This doesn't always happen, it seems
>> to be random and I can't determine what's causing it.
>>
>> For example, if Init() calls function1(), which calls function2(),
>> which calls function3(), and a breakpoint is placed in function3(),
>> then when the debugger pauses the program the stack trace only shows
>> function3() and function2(), and nothing under that.
>>
>> My setup is stock RTEMS 4.10.2, with GDB 7.2 running under MinGW
>> Windows, compiled for SPARC LEON3. The tools are prebuilt and obtained
>> from the RTEMS website.
>>
>
>You might want to try with gdb-6.4, which is the last version that work for
me.
>Later gdb versions have changed something in the SPARC stack unwinding,
>leading to all sorts of problems ...
>
>Jiri.
>
>> I've been experiencing this issue for a while now, and I've simply
>> worked around it, but it's starting to get difficult now. I was hoping
>> that someone else in the mailing list has experienced this as well and
>> knows a way around it.
>>
>> Thanks,
>> Mohammed Khoory
>>
>> _______________________________________________
>> rtems-users mailing list
>> rtems-users at rtems.org
>> http://www.rtems.org/mailman/listinfo/rtems-users
>>