NetBSD/distrib/sets
2003-05-04 00:54:41 +00:00
..
lists More wscons fonts. 2003-05-04 00:54:41 +00:00
checkflist X11 directories and files in "/etc". 2003-03-01 07:31:35 +00:00
getdirs.awk getdirs.awk: 2002-05-19 13:32:44 +00:00
join.awk getdirs.awk: 2002-05-19 13:32:44 +00:00
listpkgs set MAKE="${MAKE:-make} -j 1" to prevent problems where the output 2002-07-23 09:02:45 +00:00
Makefile In "sets" target, ensure ${RELEASEDIR}/${MACHINE}/binary/sets exists. 2003-01-30 12:16:18 +00:00
makeflist Do not try to build LKMs on evbppc, because we cannot predetermine 2003-03-15 22:32:07 +00:00
makeobsolete etc/postinstall 2003-02-20 11:40:28 +00:00
makeplist set MAKE="${MAKE:-make} -j 1" to prevent problems where the output 2002-07-23 09:02:45 +00:00
makesrctars use gzip -9 instead of --best 2002-04-12 07:08:37 +00:00
makesums improve description of -t 2003-01-03 15:26:32 +00:00
maketars Temporarily revert last. It now screws over symlinks from the destdir, 2003-04-07 17:22:19 +00:00
README

# $NetBSD: README,v 1.5 2001/10/20 06:11:21 jmc Exp $

To: "NetBSD Port Maintainers"
Subject: distribution sets, etc...
Date: Mon, 18 Jul 1994 05:02:11 -0400
From: "Chris G. Demetriou"

[ this is also going in a README in the 'sets' directory... ]

well, i've finally sat down and started doing distribution set
stuff in a "nice, machine-independent way."  Should have done
it months ago, but too much other stuff going on then...

anyway, grab ~cgd/sets.tar.gz -- it's three scripts, and a bunch
of files.

the scripts should be run from the directory where they reside.

makeflist:	output the list of files that should be in a
		distribution, according to the contents of the
		'lists' directory.

checkflist:	check the file list (as internally generated
		by makeflist) against the tree living in $DESTDIR.
		(that tree should be made with 'make distribution'.)

maketars:	make tarballs of the various sets in the distribution,
		based on the contents of the lists, the tree in
		$DESTDIR, and put the tarballs in $RELEASEDIR.
		Note that this script _doesn't_ create the 'secr'
		distribution, because (for now) it requires
		manual intervention to get the binaries right...
		(i'll add another script to create that dist, later.)

what's in 'lists':

lists describing file sets.  There are two sets of lists per file
set: machine dependent and machine-independent files. (there's
also another file in the 'man' dir, which is used by the 'man'
and 'misc' sets, but that's explained later.)

There is one machine-independent file, named "mi".  There are
N machine-dependent files (one per architecture), named "md.${ARCH}".

XXX - Temporarily (while USE_NEW_TOOLCHAIN is in effect) there are also tc.mi
and tc.old files in some of the set lists. tc.mi lists files in the new 
toolchain that didn't exist in the old toolchain. tc.old lists files removed
and aren't present in the new toolchain. All other toolchain information was
left for now in the various mi/md files. The makeflist script will check
for USE_NEW_TOOLCHAIN and pull the appropriate one into place. Once all
ports are converted to the new toolchain the entire toolchain should be pulled
into tc.mi and/or tc.${MACHINE} and everything in tc.old moved to obsolete.

the sets are as follows:

	base:	the base binary set.  excludes everything described
		below.

	comp:	compiler tools.  All of the tools relating to C, C++,
		and FORTRAN (yes, there are two!) that are in the
		tree.  This includes includes, the linker, tool chain,
		and the .a versions of the libraries.  (obviously,
		base includes ldd, ld.so, and the shared versions.
		base also includes 'cpp', because that's used by X11.)
		includes the man pages for all the binaries contained
		within.  Also, includes all library and system call
		manual pages.

	etc:	/etc, and associated files (/var/cron/tabs, /root,
		etc.).  things that shouldn't be blindly reinstalled
		on an upgrade.

	games:	the games and their man pages.

	man:	all of the man pages for the system, except those
		listed elsewhere (e.g. in comp, games, misc, text).
		Includes machine-dependent man pages for this CPU.

	misc:	share/dict, share/doc, and the machine-dependent
		man pages for other cpu's which happen to always
		be installed.

	text:	text processing tools.  groff and all of its friends.
		includes man pages for all bins contained within.


as noted, in addition to the "standard" files in each dir, there's
a file called 'md_share' in lists/man.  it's the list of man pages
that are installed from /usr/src/share, which are machine-dependent.
(note that ones that are installed from elsewhere, and thus are
installed on only one architecture, are listed in the md.${ARCH}
file.)  basically, it's grepped through, to see which of the
machine-dependent man pages that are always installed should go
in the 'man' set, and which should go into the 'misc' set.


I've made attempts to get the md files for the amiga, i386, hp300,
pc532, and sparc done.  I've not attempted the rest (though there
are empty files for all architectures except the da30, in the
various directories).  I think i got the amiga, i386, and sparc
done correctly, but i know that (because of weirdnesses in how
their last snapshots were built) i didn't get hp300 or pc532
done right.

what you folks need to do:

run 'checkflist' on a tree that you've created with 'make
distribution'.  send me the output, and i'll take care of putting
stuff on the right lists for you.  In the future, you'll get
to do it yourselves, as you add programs and include files, but
i'd prefer to do it the first time (unless you really want to do it
yourself -- in that case, please ask me to check your suggestions
re: the appropriate sets for things...)  There should be a minimal
list output by checkflist -- it should all be things that
_should_ be on your lsits, but aren't, i.e. machine-dependent
binaries, man pages, and include files.

Also, please look through the lists, and tell me if you think
i've put anything in the wrong place.  i think i've
taken care of all of the machine-dependence bugaboos, by
doing it for the N architectures, but i'm not _sure_...  8-)


thanks,

chris