FYI, I've just posted bmake-20051105.tar.gz which addresses most of 
the issues brought to my attention recently.

1/ _PATH_DEFSYSPATH defaults to just ${prefix}/share/mk
2/ boot-strap can be run with --mksrc none to skip looking for
   install-mk.  In this case though ${prefix}/share/mk/sys.mk had
   better exist
3/ boot-strap will set LIBC="" to avoid *.mk pointing it into the
   weeds.  I've also posted mk-20051104.tar.gz which removes 99% of
   the LIB*= settings - one can now put these in an optional libnames.mk if
   desired. 
4/ the screwup in job.c is fixed.

Note by doing

cd bmake
./boot-strap --objdir . [other args]

there is no reliance on VPATH working.

One issue I haven't found a fix for... on IRIX64, the *.mk in pkgsrc
insist on MACHINE_ARCH being set to either 'mipseb' or 'mipsel'.
I've no idea which if any is prefered and right now 
--with-machine-arch=mipseb
doesn't help as I expect bmake is getting 'mips' from uname(3).
Might need to introduce a --with-force-machine-arch...

--sjg

On Wed, 2 Nov 2005 13:26:06 -0500 (EST), Todd Vierling writes:
>On Wed, 2 Nov 2005, Todd Vierling wrote:
>
>> OK, so MAKESYSPATH doesn't work as an override.  If I change (in
>> makefile.boot.in):
>>
>> 	${BMAKE_ENV} ./bmake.boot -f Makefile
>>
>> to:
>>
>> 	${BMAKE_ENV} ./bmake.boot -m ${MK} -f Makefile
>>
>> it manages to start building the non-bootstrap bmake.
>
>I forgot to mention:  ...but the resultant non-bootstrap bmake still fails
>at runtime, as the compiled-in system path causes the read-'em-all behavior
>I noted below.
>
>-- 
>-- Todd Vierling <tv@duh.org> <tv@pobox.com> <todd@vierling.name>