> 	* be split into multiple MACHINEs, leaving MACHINE_ARCH as
> 	  arm32 and having common arm32 code shared, but otherwise
> 	  be separate per type/family of machine that is supported.
...
> I'd say that the ARM should do the latter, like mips, m68k, powerpc,
> and sh3 have done.

Um, having just done 1.4.2 builds for arm32, next68k, and sun3, I'd
like to complain about that.

On arm32, it took a while to build all the various kernels and ramdisk
install kernels (still faster than all my m68k's put together, though).
Also I had to remember not to boot GENERIC on either of my machines,
a CATS and a SHARK. Annoying? Yeah.

But on next68k and sun3 I ended up building the entire distribution
on two different machines, only to have large chunks of them chucked
by symlink merges on the FTP site. AFAIK no analysis is done to actually
prove that the binaries were in fact equivalent; we're just relying
on proof by construction based on a design we believe is still clean.

For all I know, there could be small port dependent differences creeping
into the m68k ports, and screwing things up. There's no established
procedure for using the common bits to speed up builds on slower m68k's;
this would be the best way for me to test them on my build machines before
a user gets that combination of bits installed on theirs. As things stand
right now, m68k users are getting a bit combination that I have not tested,
and that the people who had early access to my build products did not test.

Right now I have my CATS system build all of arm32, and the SHARK is a
stable server. If you break up arm32, then I have to use both of them for
builds (much of which will be wasted), and it will become nontrivial for
me to build A7000 or RISCPC or VOYAGER. It would help if we could somehow
develop a method for doing cross-builds in the reduced case where host and
target have the same MACHINE_ARCH.

So to me, splitting up arm32 means more bits being crunched per release,
with negative real benefit. It would not be hard to make sysinst DTRT
when installing kernels on arm32 systems.

I would be happier if the m68k ports moved closer together, actually.
Due to the sun3+sun3x merger, I was able to build both of those on a
single sun3/60.

Todd Whitesel
toddpw @ best.com