Re: [avr-libc-dev] reconf

[Russell Shaw]

Russell Shaw wrote:
> E. Weddington wrote:
>
> > Paul Schlie wrote:
> >
> > > > From: "E. Weddington" <address@hidden>
> > > > Joerg,
> > > >
> > > > Do you think the version checking in the reconf script is strictly
> > > > necessary?
> > > > How do you feel about it's removal?
> > >
> > > Personally wouldn't mind if all the reconf/doXXX scripts
> > > disappeared in favor of an  updated configure/make.
> > >
> > > (ideally using more recent autoXXX scripts if
> > > someone can figure how to do it construct them)
> > Thank you for your opinion. Do you have a patch?
>
>
> I looked at converting it to autoconf 2.59, but this
> line in configure.in is the hard part:
>
>   AC_OUTPUT(
>     ...
>     [if test -n "$CONFIG_FILES"; then
>       . ${avrlib_basedir}/config-ml.in
>     fi],
>     ...
>   )
>
> config-ml is a big shell script dealing with multilibs.
> I assume multilib was a cygnus thing for building the
> libraries for *all* cpus. config-ml is full of cpus
> except avr. It also says:
>
> # We have to handle being invoked by both Cygnus configure and Autoconf.
>
> So, being avr specific, should all this ancient multilib complication
> be deleted?

I found multilibs is to build multiple versions of the same library
for the same cpu, but with different parameters such as with/without
soft floating point etc. Maybe the newlib method could be used to
sort between the 'half' enhanced avrs.

> The other way is to save all the effort for putting avr-libc
> into newlib or something.