Re: avr and frame unwinding

[Andrew Cagney <ac131313 at redhat dot com>]

> :):)Is this arrithmetic correct - I understand the ``* 2'' but not the ``>>8''.
> :):)
> :):)>               pc = (extract_unsigned_integer (buf, 2) * 2) >> 8;
> :)
> :)That's the ugly part I don't understand. It seems to give the correct
> :)result, but now that I think about it more, it could mean that my memory
> :)address is off by 1. I will have to re-examine that.
> :)
> :):)
> :):)this memcpy will need to be a
> :):)
> :):)	store_unsigned_integer (bufferp, pc, SIZEOF_AVR_PC);
> :):)
> :)
> :)I tried that, but it performed a endian byte swap and the PC came out wrong.
> :)I dug around and saw what looked to be too many byte swaps. 
> :)
> :):)>               memcpy (bufferp, &pc, sizeof(pc));
> :):)>             }
> :):)>           else
> :):)>             {
> :):)>               read_memory (this_saved_regs[regnum], bufferp,
> :):)>                            register_size (current_gdbarch, regnum));
> :):)>             }
>
> I think found the root of the ugliness. When the avr performs a call 
> instruction the PC is pushed onto the stack, but it turns out that it is 
> pushed in big endian order. For the most part though, the avr is little 
> endian.

That would explain it.  I guess it needs an explicit big endian extract 
followed by a little endian store (via store_unsigned_integer).

Andrew