gdb && Eclipse && Harvard cpu

[Usachev M.]

I use avr-gdb 6.5 && simulavr (latest WinAVR package). When C-program contains static variables, gdb works fine (requests locals and variables led to generation packets like $m800060).
But if I use stack local variables, command "x my_var" led to the packet like $m60 (not $m800060).
And "x *&my_var" works fine - "$m800060").
why?

I use Eclipse (not command line) where disassemble and inspecting locals and variables are automatic, so it is important for me to correct this problem.

[Weddington, Eric-2]

 

> -----Original Message-----
> From: Usachev M. [mailto:um.lan@...]
> Sent: Tuesday, December 25, 2007 2:00 AM
> To: gdb@...
> Subject: gdb && Eclipse && Harvard cpu
>
>
> I use avr-gdb 6.5 && simulavr (latest WinAVR package). When C-program
> contains static variables, gdb works fine (requests locals
> and variables led
> to generation packets like $m800060).
> But if I use stack local variables, command "x my_var" led to
> the packet
> like $m60 (not $m800060).
> And "x *&my_var" works fine - "$m800060").
> why?
>
> I use Eclipse (not command line) where disassemble and
> inspecting locals and
> variables are automatic, so it is important for me to correct
> this problem.

In talking to Joerg Wunsch (co-maintainer of the AVR toolchain, and who
at least uses GDB), he said:
"Sounds like a genuine bug to me.  Personally, I never used the "x"
command to print the contents of local variables, but rather the
"p[rint]" command instead."

I know that this may not be much help to you at the moment.

Please fill in a GDB bug report for this for the AVR target, and send me
the bug # so I can track it.

Thanks,
Eric Weddington
Product Manager
Atmel

[Usachev M.]

This problem exists for version 6.5 avr-gdb and avr-insight, included into
may2007 package of winavr. Yesterday I have tried december2007 release of
winavr, in which gdb 6.6 is included - it works fine.

It seems the problem is solved.