RE: [avr-libc-dev] [RFC] Unified ELF file

[Eric Weddington]

> -----Original Message-----
> From: Paulo Marques [mailto:address@hidden
> Sent: Monday, September 24, 2007 5:17 AM
> To: Eric Weddington
> Cc: avr-libc-dev; 'Erik Christiansen'
> Subject: Re: [avr-libc-dev] [RFC] Unified ELF file
>
> Eric Weddington wrote:
> > [...]
> > Some questions:
> > - Does this hybrid approach have any merit? Should it be
> done differently?
>
> I for one would be ok with it, even in its current form ;)

That's good to know! :-)

> > - How can the API be changed to ensure that all fuse bytes
> are set by the
> > user?
>
> Couldn't this be done using "__attribute__((weak))"? We could declare
> these variables in some core code (crt0?) and let the linker
> toss them
> away if the user declares them again in the code.

There's an equivalent assembler directive. As this would be changing the
startup code module which is written in assembly, we would use the assembler
directive.

>
> > - Any thoughts on why -combine -fwhole-program is failing
> to include all
> > input fuse sections?
>
> I think gcc is able to tell that those variables aren't being used
> anywhere in the program so it just optimizes them away.
>

Then the .lfuse section should be optimized away too and its not.

Eric Weddington