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

[Eric Weddington]

> -----Original Message-----
> From:
> address@hidden
> [mailto:address@hidden
> org] On Behalf Of Michael Hennebry
> Sent: Wednesday, September 26, 2007 9:52 PM
> To: Dave Hylands
> Cc: avr-libc-dev
> Subject: Re: [avr-libc-dev] Re: [RFC] Unified ELF file
>
> On Wed, 26 Sep 2007, Dave Hylands wrote:
>
> > On 9/26/07, Colin O'Flynn <address@hidden> wrote:
> > > Hi,
> > >
> > > > I propose a counter-argument: it makes no sense for
> *the user* to want to
> > > > program a single fuse byte.
> > >
> > > I agree with Eric 100% here. If you are going to be
> setting the fuse bytes, you
> > > *must* set them all. Otherwise you are assuming the
> current state of fuse
> > > bytes. Sure they *should* be default, but if someone else
> had their hands on
> > > the chip it might change. That's the sorta scenario where
> two months down the
> > > road stuff stops working, and you can't figure out why...
> >
> > Perhaps, what should be included is a set of fuse bytes and a
> > corresponding mask. This would allow the program to set
> just the bits
> > it was interested in, rather than being required to set all of the
> > bits.
>
> He might not do it because it allows
> something he wants to forbid.
>
> How about eliminating the surprise issue this way:
> Include the masks in the elf file.
> Declare that a well-made tool will not accept a mask
> other than all-ones unless explicitly told to.
> A bondage and discipline tool will not
> accept a mask other than all-ones.
>
> Whatever is done, don't use default fuse values.
> Default fuse values will produce surprises.

Oh? And how would non-default fuse value not produce surprises? Default fuse
values are known good values and can be looked up in the datasheet.

I have many real-world cases where this overall feature is useful, and if
the user acceidentally does not specify a fuse byte, then a default should
be used. If it can even be done.

I'm not turning this into a bike-shed event. Patches welcome.

Eric