Remove support for the historic Intel 80386

[Chris Johns]

On 27/04/2018 08:47, Joel Sherrill wrote:
> On Wed, Apr 25, 2018 at 9:34 PM, Chris Johns <chrisj at rtems.org
> <mailto:chrisj at rtems.org>> wrote:
>     On 25/04/2018 23:48, Joel Sherrill wrote:
>     > 
>     > To answer your m68k question...I can't find the post to gcc@ but there are only
>     > a handful
>     > of ports using the cc and that is a barrier to future changes. The m68k, v850,
>     > and a few
>     > others are at risk until the port is updated to use more modern gcc internals. 
>     > It was
>     > discussed that if these ports are not updated, they will be deprecated.
>     > 
> 
>     https://gcc.gnu.org/ml/gcc/2018-01/msg00089.html
>     <https://gcc.gnu.org/ml/gcc/2018-01/msg00089.html>
> 
> Thanks! We need to watch this topic because the mc68040 is more important
> than the mc68060 as a cutoff point. We may also want to consider what to
> do about 6830x and cpu32 variants. If we want those to stay, that may increase
> the cores needed a lot. That's effectively keeping the 68000 and 68020.

If the Coldfire and some m68k variants are import and we would like to maintain
a current gcc on those devices then someone will have to migrate the
architecture off cc0 and then to LRA. This may be directly with patches to gcc
or via funding. I think it is only a matter of time before our requests are
respectfully considered and rejected.

We may need to be more pragmatic about how we move to the latest gcc. We may
arrive at a point in time where we sit on a older version for architectures we
want to support that gcc has dropped. I accept this is a change in our current
policy and could result in functionality on those architecture being limited in
some way.

Chris