I just did some testing (redirecting all the paths to a scratch dir)
and found that the bootstrap completes on MacOS X 10.4.3, but on 10.3.9
and 10.2.8, bmake fails its unit tests, as first noted by Mark Perkins
in:

    http://mail-index.netbsd.org/tech-pkg/2005/11/06/0016.html

and which I reiterated in:

    http://mail-index.netbsd.org/tech-pkg/2005/12/10/0015.html


The 10.4.3 system is using XCode Tools 2.2 w/GCC 4.0.1.  The 10.3.9
system is using XCode Tools 1.5 with GCC 3.3 and the November 2004 GCC
update.

The 10.2.8 system is using the December 2002 Developer tools with the
August 2003 GCC updater.  The gcc 3.1 compiler is selected due to
stability issues.  (The machine has an MPC604 CPU running 10.2.8 by way
of XPostFacto.)

The 10.2.8 has an existing bootstrap built quite some time ago.  I just
attempted to build devel/bmake and it failed in the same fashion.


I've been going back through the archived messages on this subject and
it appears that, on these systems, bootstrapping bmake separately (outside
of pkgsrc/bootstrap or devel/bmake) works?  How can this be when to be
useful in pkgsrc, one should specify the same arguments to the "boot-strap"
script that are passed in the bootstrap/bootstrap script or devel/bmake's
Makefile?

At present, to get bootstrap pkgsrc on MacOS X 10.2.8 or 10.3.9, it
appears that libnbcompat gets built and installed through the
bootstrap/bootstrap script, then bmake needs to be built somewhere and
installed semi-manually, then the rest of the tools could be built
through pkgsrc?

Any clues on what changed in bmake that is causing it to not be built
correctly through the bootstrap process on MacOS X 10.2.8 and 10.3.9?

-- 
John D. Baker, KN5UKS                    NetBSD     Darwin/MacOS X
jdbaker(at)mylinuxisp(dot)com                 OpenBSD            FreeBSD
BSD -- It just sits there and _works_!
GPG fingerprint:  D703 4A7E 479F 63F8 D3F4  BD99 9572 8F23 E4AD 1645