[avr-libc-dev] Re: avr-libc-dev archives

[E. Weddington]

On 17 Dec 2002 at 22:31, Joerg Wunsch wrote:

> As E. Weddington wrote:
> 
> So yes, i've seen that FAQ entry submission.  I have to think a bit
> about it.  It seems all you gonna do is to create a bootloader.  So
> why don't you simply relocate the entire .text segment?

The issue is not creating a bootloader. It's the FAQ of how to 
relocate code to a fixed address. The bootloader stuff is just an 
example.

There's currently an unanswered question on avr-gcc-list that fits 
this.

> Normally, no program is going to provide the code for both, the
> bootloader and the regular application, within a single ELF file.

I do. The bootloader code is part of the regular application. This 
will be for upgrading the software when the unit is in the field. If 
you recall on this list, I posted an RFC about adding a file to 
avrlibc, avr-boot.h, which provides a low-level API to read / write 
the flash, clearing the RWW section, etc. It's been tested on a 128, 
i.e. it uses RAMPZ to access the entire 128 memory. Several people 
have asked for it and have used it with success. The API is 
essentially inline assembly macros.

> Arguably there should be an option to avoid the interrupt vectors
> entirely for this.  A friend of mine used to do this before with a
> privately hacked linker script (and since this friend is supplying me
> with AVR test boards as i desire, i'm willing to support him as good
> as i can ;-).  It needs a few more thoughts about how to make this an
> official option.   OTOH, now that the ATmega128 offers the ability to
> have a second array of interrupt vectors in the bootloader section,
> there's even less pressure to really have it. ;-) -- J"org Wunsch             
>                         
>       Unix support engineer address@hidden       
> http://www.interface-systems.de/~j/