[1161] in NetBSD-Development
Re: Installation changes and plans
daemon@ATHENA.MIT.EDU (Greg Hudson)
Sun Dec 17 18:57:36 1995
Date: Sun, 17 Dec 1995 18:57:08 -0500
From: Greg Hudson <ghudson@MIT.EDU>
To: Greg Hudson <ghudson@MIT.EDU>
Cc: netbsd-dev@MIT.EDU
In-Reply-To: "[1158] in NetBSD-Development"
> * I move the installation source out of the NetBSD source tree
> and into /mit/netbsd/dev/install. This step will probably
> happen tonight.
Done. I still need to modify the Makefiles to use crunchgen and
crunchide from some place other than the default path.
> * I examine all of the changes we've made to the
> netbsd-current source tree, and catalog and propagate the
> ones that need to be applied to the netbsd-1.1 tree.
I've examined all of the changes we've made. I came up with the
following sets of changes which may need to be propagated:
* There is an AFS bug which prevents the stock install from
working (if you fchown a file non-writable before closing
it, the file doesn't get written out). It may be difficult
to fix the AFS bug, in which case we may need to hack
install again.
* We distribute kernels with PCVT, which requires some changes
to /dev/MAKEDEV and /etc/ttys. We could install GENERIC
kernels instead of ATHENA kernels when someone doesn't do
the athena installation, and make the ttys and MAKDEV
changes part of the Athena installation.
* To support getting rid of the "Insert root filesystem
floppy" message, a change to sys/arch/i386/boot/boot.c is
necessary.
* sys/compat/common/Makefile{,.inc} and sys/lib/libkern{,.inc}
don't work properly if compile is not physically located in
sys/arch/i386/compile. Likewise,
sys/arch/i386/Makefile.i386 needs an absolute path for S=
if compile is not located there. I have changes to
sys/compat/common and sys/lib/libkern that probably ought to
be propagated to the repository; they probably also ought to
go into our 1.1 tree.
* I have a hack to make to support path-canonicalization of
filenames using the :C variable expansion option. We were
using this for proper operation of "make obj" when /usr/src
is a symlink.
* The kernel config files for {ATHENA,INSTALL}-{AHA,BT}, of
course.
* At jhawk's recommendation, we currently install kernfs and
procfs on people's systems; we have a hacked
etc/mtree/4.4BSD.dist to this end. Maybe this should be
part of the Athena installation.
* ~root/.klogin contains a dummy entry, which jhawk finds
objectionable.
* Resolving the symlinks issue the way we did before requires
changes to 30 files. If we're willing to tolerate the
symlinks being to absolute paths, that cuts out a lot of
modifications, but jhawk doesn't like that answer (and I
don't like it either, really).
I may resolve this issue by making copies instead of
symlinks when files aren't the same directory; it won't
increase the size of the operating system very much.
* We have changes to config and to files in /usr/share/mk to
allow users to build out of our source tree. I may abandon
that for the 1.1 release (or do just the kernel part of it
by hacking config and having people run the modified config
out of the netbsd locker), and have people copy the source
tree if they want to build out of it.
I'll take care of these changes to the 1.1 tree over the next few
days. In addition to the above changes, we already have John Kohl's
symbol-loading hacks incorporated into the 1.1 tree. I'm not sure
whether they should stay there or not.
