Patch fixing 3.3 bug PR 9745 and PR 10021

[Jim Wilson]

On Tue, 2003-07-08 at 21:37, David Edelsohn wrote:
> 	Who are "those people"?  As you stated later in your message:

Please don't turn this into an argument.  I am trying to make
constructive progress on a difficult issue, while lacking access to
information, hardware, and software necessary to fully understand it.  I
am not trying to insult anyone in the process, I am just trying to fix
the problem.

The context should have made it clear that I was referring to people
complaining about the performance problems caused by my earlier patch. 
Obviously, everyone cares about performance including me.  However, I
have a very real immediate problem here.  I have seen no direct evidence
of this performance problem, just your claim that there is one.  I
believe your claim, but it would be nicer if I had facts and figures,
e.g. SPEC results showing which benchmarks have performance regressions
and by how much.  Then I would be better able to understand whether this
is a significant performance regression or one we can live with.  If you
could provide these figures that would be constructive.  It would also
be constructive if IBM or Apple gave me free hardware so I could produce
the figures myself.  But just arguing that everyone cares about
performance is not constructive.

> 	According to the GCC Development Plan, which describes
> responsibility for fixing bugs, David Miller is reponsible for fixing the
> latent bug that his patch exposed.

This is not constructive.  David Miller's patch was a useful gcc
reorganization.  It is unavoidable that major useful changes will
require some later bug fixing.  If we require people contributing major
useful changes to fix every bug caused by that patch on demand until the
end of time, then no one would ever be able to make any major change
because no one can realistically make that kind of commitment.  There
should be a statute of limitations to this clause, and David Miller's
patch should be considered past the statute of limitations because it
was over a year ago, thus we should not be requiring David to fix this
problem.

> I will test the patch.
>        I believe that restoring the original behavior is the best we
>can do without a complete fix for the alias.c problems.

Thanks.  I believe this patch is likely to be a good one also.  And in a
followup message you mentioned that the patch solved 9745 so this is
very good news indeed.

Jim