Debugging question

Till Straumann strauman at slac.stanford.edu
Thu Feb 24 23:35:23 UTC 2005


Peter Dufault wrote:

> Can I ask you directly what debugging setup I should setup for the 
> mvme5500, i.e., GDB serial?

AFAIK, there is no support for GDB on this BSP out of the box. In cases 
Iike yours,
i.e., when hitting an exception you can trace back manually:

a) disassemble the ELF object (powerpc-rtems-objdump -dr <objfile>)
b) work your way through the stack / register dump

With a cross-gdb [powerpc-eabi target] you can do this symbolically, of 
course,
i.e., gdb will find the source code lines corresponding to the PC 
addresses on the
stack [list *<pc_address>].

Regarding 'live debugging':

I wrote a GDB stub that can be used on powerpc/shared + derived BSPs. 
Unfortunately, I had
no time to write documentation yet but the software as such is somewhat 
working.

Here's what you need:

a) a recent [i.e., yet unreleased] version of 'libbspExt'. This library 
provides a simple API
    to hook/register exception handlers. The GDB stub needs this to 
register its handler
    which intercepts exceptions.

b) rtems-gdb-stub software package. This provides a GDB helper thread 
implementing
    the stub [interpreter of the GDB remote protocol]. Should work on 
sockets and serial
    ports but serial port support is not implemented yet [not hard to 
do]. This means that
    you need a working TCP/IP stack.

   The package consists of generic material and architecture specific 
pieces which are
   currently implemented only for PPC/shared and pc386.

c) patch for GDB (small bugfix [i.e. fix is small, implications of the 
bug are big]
   + support for loadable modules on the target [not an issue if you 
don't use the cexp shell])

If you want to experiment with this stuff let me know and send you the 
necessary bits...

HTH
-- Till

>
>
> I will research these things myself, but if you're willing to answer 
> then I'm willing to ask.  I'll be sure to document what I figure out 
> on the mailing list and rtems "wiki".
>
> Thanks,
>
> Peter
>
> Peter Dufault
> HD Associates, Inc.
>





More information about the users mailing list