Get rid of: Previous frame identical to this frame (corrupt stack?) when using kernel-level stepping

[Emmanuel Michon <emmanuel dot michon at polytechnique dot org>]

Hello,

I'm almost exclusively using gdb 6.8 and sde-gdb as a cross debugger
to a MIPS JTAG interface.

This comes very handy to debug interrupt vectors or assembly code, and
there are many relevant cases where the stack consistency `a la mips' is
not enforced all the time:

* when you're stepping in the code that sets up the stack,
* when you run simple o32 compliant LEAF routines that use a0..a3 they
don't need
$sp to be setup,
* when you run FORTH or stuff that just uses registers a different way.

If `Previous frame identical...' message was just a warning I wouldn't care,
but it prevents the code that changes $pc to the location you want to
complete (set $pc,
or load, etc).

In my daily experience e.g. JTAG-level debugging, symbolic information
(proper stack display,
displaying char* sometimes following bad pointers) is often a
nuisance, but accurate `step instruction'
and breakpoint a must.

I tried stuff like set trust-frame-pointer 1,
set language assembly,
but could never work this around.

Am I wrong in what I try to do?

Do you have a suggestion?

-- 
Très sincèrement,

e.m.