Subject: Re: Seperate share tarball?
To: None <tech-install@netbsd.org>
From: Simon Burge <simonb@netbsd.org>
List: tech-install
Date: 06/17/1999 16:44:09
I've done a little further work on a UST (unified share tarball), and
here's the results if a "diff -qr" of the 1.4 pmax and i386 usr/share,
ignoring the manpages that I recently made sure get installed on all
ports:

	Files i386/usr/share/games/cards.pck and pmax/usr/share/games/cards.pck differ
	Files i386/usr/share/info/binutils.info and pmax/usr/share/info/binutils.info differ
	Files i386/usr/share/info/dc.info and pmax/usr/share/info/dc.info differ
	Files i386/usr/share/info/gdb.info and pmax/usr/share/info/gdb.info differ
	Files i386/usr/share/info/gprof.info and pmax/usr/share/info/gprof.info differ
	Files i386/usr/share/info/uucp.info and pmax/usr/share/info/uucp.info differ
	Files i386/usr/share/man/cat1/ld.0 and pmax/usr/share/man/cat1/ld.0 differ
	Files i386/usr/share/man/man1/ld.1 and pmax/usr/share/man/man1/ld.1 differ
	Files i386/usr/share/man/whatis.db and pmax/usr/share/man/whatis.db differ
	Only in i386/usr/share: pcvt

and here's a diff between pmax and sparc (something big endian):

	Files pmax/usr/share/games/cards.pck and sparc/usr/share/games/cards.pck differ
	Files pmax/usr/share/games/fortune/fortunes-o.dat and sparc/usr/share/games/fortune/fortunes-o.dat differ
	Files pmax/usr/share/games/fortune/fortunes.dat and sparc/usr/share/games/fortune/fortunes.dat differ
	Files pmax/usr/share/games/fortune/fortunes2-o.dat and sparc/usr/share/games/fortune/fortunes2-o.dat differ
	Files pmax/usr/share/games/fortune/fortunes2.dat and sparc/usr/share/games/fortune/fortunes2.dat differ
	Files pmax/usr/share/games/fortune/limerick.dat and sparc/usr/share/games/fortune/limerick.dat differ
	Files pmax/usr/share/games/fortune/startrek.dat and sparc/usr/share/games/fortune/startrek.dat differ
	Files pmax/usr/share/games/fortune/zippy.dat and sparc/usr/share/games/fortune/zippy.dat differ
	Files pmax/usr/share/man/cat1/ld.0 and sparc/usr/share/man/cat1/ld.0 differ
	Files pmax/usr/share/man/man1/ld.1 and sparc/usr/share/man/man1/ld.1 differ
	Files pmax/usr/share/man/whatis.db and sparc/usr/share/man/whatis.db differ
	Files pmax/usr/share/misc/termcap.db and sparc/usr/share/misc/termcap.db differ
	Files pmax/usr/share/misc/vgrindefs.db and sparc/usr/share/misc/vgrindefs.db differ

and another between pmax and alpha (32bit vs. 64 bit):

	Files pmax/usr/share/games/cards.pck and alpha/usr/share/games/cards.pck differ
	Files pmax/usr/share/games/fortune/fortunes-o.dat and alpha/usr/share/games/fortune/fortunes-o.dat differ
	Files pmax/usr/share/games/fortune/fortunes.dat and alpha/usr/share/games/fortune/fortunes.dat differ
	Files pmax/usr/share/games/fortune/fortunes2-o.dat and alpha/usr/share/games/fortune/fortunes2-o.dat differ
	Files pmax/usr/share/games/fortune/fortunes2.dat and alpha/usr/share/games/fortune/fortunes2.dat differ
	Files pmax/usr/share/games/fortune/limerick.dat and alpha/usr/share/games/fortune/limerick.dat differ
	Files pmax/usr/share/games/fortune/startrek.dat and alpha/usr/share/games/fortune/startrek.dat differ
	Files pmax/usr/share/games/fortune/zippy.dat and alpha/usr/share/games/fortune/zippy.dat differ
	Files pmax/usr/share/man/whatis.db and alpha/usr/share/man/whatis.db differ
	Files pmax/usr/share/misc/termcap.db and alpha/usr/share/misc/termcap.db differ


pcvt is on the known list of something that's not MI, and as per Todd's
suggestion I'll have a look at what needs to be done to move this and
the keymaps to usr/libdata.  Without looking, I'd guess that these (the
keymaps) should be at least architecture independant so that an amiga,
an atari and an x68k can share /usr.


I dunno what's going on with the i386 info files.  I'm guessing
that something just went wrong with the i386 build.  The header of each
info file is different:

	--- pmax/usr/share/info/binutils.info   Fri May  7 18:20:23 1999
	+++ i386/usr/share/info/binutils.info   Mon May 10 04:21:05 1999
	@@ -1,5 +1,6 @@
	-This is Info file binutils.info, produced by Makeinfo version 1.68 from
	-the input file
	+This is Info file
	+/usr/src/gnu/usr.bin/binutils/../../dist/binutils/binutils.info,
	+produced by Makeinfo version 1.68 from the input file
	 /usr/src/gnu/usr.bin/binutils/../../dist/binutils/binutils.texi.
	  
	 START-INFO-DIR-ENTRY

and it looks like a lot of offsets are different (maybe because of the
different header) but the rest of the info files seem identical.


The ld(1) manpages is because some ports use ld (only the a.out ones?)
and others ld.new.  I'd guess that everything is not going to be moved
to ELF for 1.5, so we need a way to handle this.


Fortune seems to think it uses network byte order, but the files are
different across different endian machines, and also different between
32bit and 64bit little endian machines (pmax vs. alpha).  A pmax unstr
(built from games/fortune/unstr) can't read a sparc data file.  I'll
have a look at this one soon.


The Berkeley DB files could also be fun.  Again since the alpha and pmax
are different, but the pmax and i386 are the same there's probably some
64 bit issues.  If these are easy to fix, the next question would be
which endianness would we ship?  I can't see a way of telling dbopen()
to tell it to build a particular endianness file.


That leaves the monopoly cards, which I haven't looked at yet.  It
shouldn't be too hard to come up with a network byte order solution for
that one either.


Hmm, not there yet with usr/share...  xfont.tgz will be next on the
hitlist :-)

Simon.