Re: rearranging REG_ALLOC_ORDER in gcc/config/rs6000/rs6000.h

[Jim Wilson <wilson at tuliptree dot org>]

In case it wasn't clear, I was stating info about past discussions in
order to try to be helpful.  You left that sentence out of the part that
you quoted.

I think your message is misdirected.  I did not make any statement about
the usefulness of the feature Graeme Peterson proposed.  You seem to
have assumed I am against it.  Your assumption is in error.  I am
neither for it nor against it.  I am trying to be agnostic to avoid
being drawn into an argument, but I seem to have failed.

I think your arguments are weak.  Trying to convince someone that a
feature with bugs is a good idea is a poor strategy.  A better strategy
is to convince them that the kernel feature and the implemented feature
are two distinctly different features, each of which is useful in its
own right, and therefore the fact that the implemented feature can't be
used by a kernel is not a good reason to reject it.  Then it becomes an
issue of documenting the feature properly to avoid confusion.

If you want to be helpful, point Graeme at the discussion you started
when you proposed a similar patch.  I don't recall when you posted it,
maybe you do?

Jim