Re: [avr-libc-dev] AVR toolset critical issues

[E. Weddington]

Theodore A. Roth wrote:

> On Tue, 23 Nov 2004, E. Weddington wrote:
>
>  
>
> > E. Weddington wrote:
> >
> >    
> >
> > > Hello All!
> > >
> > > There are a number of critical issues with the AVR toolset (binutils,
> > > gcc, avr-libc) that need action soon, especially timing these with the
> > > upcomping 4.0.0 release, which seems to be targeted for early 2005
> > > (possibly even January).
> > >
> > > 1. New AVR devices.
> > > There are patches in the avr-libc Patch Manager to add these devices
> > > to the toolset: tiny13, tiny2313, at90can128, mega48, mega88, mega168.
> > > These patches are for binutils, gcc (HEAD and 3.4.x patches), and
> > > avr-libc.
> > >      
>
> Marek already committed the binutils part of this to CVS. I think Marek
> was ready to commit to gcc, but at the time he tried, he couldn't test
> it due to a broken build.
>  
HEAD is now fixed for the avr, so testing should be able to go forward.

>  
>
> > > There needs to be *new patches* created for these devices:
> > > mega165, mega325, mega3250, mega645, mega6450. But the important thing
> > > is that these patches need to get committed.
> > >      
>
> I'll see what I can do to get a patch for gcc out by wednesday night for
> review and then possible submission to gcc-patches thursday morning or
> early afternoon. I don't have much free time between now and wednesday
> evening though. :-(
>
>  
Thanks!

> > > 2. Per-device multi-lib.
> > > We've discussed this item before and there seems to be a lot of
> > > agreement that the toolset needs to get rid of the "architectures",
> > > which isn't really supported by Atmel, and go to per-device
> > > multi-libs. Ted said that Marek had some ideas about this. There's a
> > > CVS branch for it in avr-libc. But that's as far as it's gotten. Since
> > > we're adding new devices, now would be a good time to get this
> > > implemented, especially because of 4.0.0 coming up soon.
> > >      
>
> I don't see this happening for 4.0 at this point.
>
>  
That's a shame, but I understand.

> > > 3. Update avr-libc auto* tools.
> > > Unfortunately, the per-device multi-lib branch is the same as the new
> > > auto* tools branch. Perhaps they can both be done at the same time? If
> > > not, it would be great to seperate the two.
> > >
> > > 4. Warning on misspelled signal names.
> > > Ted has a GCC patch for this. If this is good to go, it really needs
> > > to get committed to GCC.
> > >      
>
> I have never heard of this being used or tested by anyone. I can post it
> to gcc-patches, but I'm not holding my breath.
>
>  
Well, you tested it didn't you? :-)

I'm planning to get this included in the next WinAVR release, so there 
should be wider testing later....

Eric