On Wed, May 31, 2006 at 01:07:09AM +0900, OBATA Akio wrote:
> On Tue, 30 May 2006 17:12:06 +0200, joerg@britannica.bec.de wrote:
> > On Tue, May 30, 2006 at 11:50:14PM +0900, OBATA Akio wrote:
> > > > bmake test in bootstrap/work/DragonFly works here very well and it
> > > > didn't attempt to read /usr/share/mk/sys.mk.
> > > 
> > > Really?
> > > Even if add something like `CC=false' to /usr/share/mk/sys.mk,
> > > or /usr/share/mk/bsd.obj.mk, bmake test works well on DragonFly?
> > 
> > Yes, I can add .error broken to /usr/share/mk/sys.mk and it just works.
> > Note that the bootstrap itself calls make, so it is natural that it
> > wouldn't work with that completely.
> 
> What are you talking about?
> Bootstrap works completely or not?
> If you just type `bmake test', it will work fine.

I can't run a full bootstrap with ".error broken" in
/usr/share/mk/sys.mk, because the system make as used by bootstrap would
not work.

> > Just to trace down where this comes from, can you ktrace the bootstrap
> > and find out when /usr/share/mk/sys.mk is accessed? It should only
> > happen by /usr/bin/make, not bmake. Also check that the _PATH_DEFSYSPATH
> > definition in the Makefile is really available.
> 
> No, Once more.
> In makefile.boot, set MAKESYSPATH environment, so bmake find not from 
> _PATH_DEFSYSPATH but from MAKESYSPATH.

I asked *two* questions.
(a) Is the _PATH_DEFSYSPATH embedded into bmake correct?
(b) *When* is /usr/share/mk/sys.mk accessed by bmake?

When the answer to (a) is yes, I have no problem with just modifying the
bootstrap script to be more defensive. But I want to ensure that it is
correct first.

Joerg