Re: [avrdude-dev] device signature string

[Jan-Hinnerk Reichert]

On Saturday 13 December 2003 00:25, Michael Mayer wrote:
> On Fri, Dec 12, 2003 at 23:55:23 +0100, Jan-Hinnerk Reichert wrote:
> > Have  you thought about using the 't' command for autodetection.
> > This should work as long as this is really a bootloader.
>
> Regarding only the butterfly no autodetection in necessary, it uses
> only the Mega169 with no alternatives. This was my starting point -
> in this special case the -p option is redundant and I proposed the
> defininition of a default device to be able to change the -p from
> mandatory to optinal in a consistent way. But using the device
> signature is a more general and flexible approach to this problem.
>
> > IMHO the reason for this is that a memory type allows you to
> > specify the command for reading. If you implement signature as
> > memory you must give a length, of course ;-(
>
> ok. So there is no easy way to get rid of the depency from struct
> avrpart.

Seems like it would be best to drop signature as a memory type.
As far as I can tell, reading signature is only implemented for 
avr910. 
Also note that there is already a function for reading signature in 
the programmer-struct.

However, I believe that we should really do a cleanup of avr.c and 
main.c first. During this cleanup we have to rethink the 
initialisation-process, anyway.

/Jan-Hinnerk