6.x from i386 to amd64

[Greg Black]

On 2006-10-31, Mark Linimon wrote:
> On Wed, Nov 01, 2006 at 06:52:27AM +1000, Greg Black wrote:
> > I found that a very large number of ports that mattered to me were marked
> > i386 only.
> 
> In some cases these designations are obsolete.  They will require people-
> power to work through them and see if they are overused.
> 
> In particular, many of these ought to have logic to set BROKEN to say
> "currently doesn't work on amd64" rather than *_FOR_ARCHS which indicates
> "can't ever work on amd64".  Even in some of those cases "currently doesn't
> work" might be obsolete; it will take people with amd64 boxes running
> native willing to test them and report back.

Thanks for the extra information; that makes the situation much
clearer.

> Yes, this is going to take a great deal of hard work by many people.

Indeed.

And that's the point I was making in my original contribution to
this thread which was begun by somebody seeking guidance on the
process for migrating from i386 to amd64 mode.  I only jumped in
to point out that such a migration could lead to pain, because
of the situation with the ports.  That's going to be an issue
for quite some time.

One thing I failed to mention at the outset, and which bears
mentioning now, is that FreeBSD-6.1 (and later) itself works
just fine on amd64, and I have no concerns about that.  In the
case of my particular workload, the performance of my box is
pretty much the same in either i386 mode or amd64 mode.  That
box shows up in dmesg as:

CPU: AMD Athlon(tm) 64 X2 Dual Core Processor 3800+ (2009.27-MHz 686-class CPU)
real memory  = 2147418112 (2047 MB)
FreeBSD/SMP: Multiprocessor System Detected: 2 CPUs

> > I didn't look for PRs and didn't submit any.
> 
> If everyone does that, then yes, the situation won't improve.

Fair enough.  In my defence, I'm fully committed at present and
I have only one amd64 machine which I need for my real work.  I
can't afford to run it in amd64 mode, because so much of what I
need is currently broken in a 64-bit world.  That much of that
broken software is interpreters and compilers for languages I
use is a pretty sad reflection on people who should know a bit
more about writing correct software, but that's got nothing to
do with FreeBSD, except as a platform for running it on.

Perhaps I'll have some more time available next year (by which
time I hope to have completed my current projects), and then I
might be able to drop out of my other free software commitments
and have a look at helping out with some of the FreeBSD ports.
But that's certainly a few months away.

Greg