[1314] in Kerberos_V5_Development
Re: sub-packages of krb5 for autoconf purposes
daemon@ATHENA.MIT.EDU (Tom Yu)
Fri Jun 14 14:03:02 1996
Date: Fri, 14 Jun 1996 14:02:48 -0400
To: "Barry Jaspan" <bjaspan@MIT.EDU>
Cc: tlyu@MIT.EDU, krbdev@MIT.EDU
From: Tom Yu <tlyu@MIT.EDU>
In-Reply-To: <9606141446.AA06941@DUN-DUN-NOODLES.MIT.EDU>
>>>>> "bjaspan" == "Barry Jaspan" <bjaspan@MIT.EDU> writes:
bjaspan> I don't quite understand how the util directory fits into
bjaspan> your list. Some thing in util (db, ss at least) are
bjaspan> needed by the KDC and admin system; others (pty) are
bjaspan> presumably needed by telnet/rlogin. Perhaps your list is
bjaspan> confused because the util directory is itself
bjaspan> confused. :-)
Most of util can be built with the rest of the tree. It's things like
appl/telnet, util/pty, and util/berk_db that really deserve separate
configure.in files so that we can distribute them as separate
packages. Marc's work on berk_db uses its own configure script. I
modify my list as follows:
util/berk_db
util/pty
appl/telnet
rest of tree
Also, the sequencing of the building of packages can be controlled.
AC_CONFIG_SUBDIRS can be used for its intended purpose, which is
configuring packages that have their own configure scripts.
Therefore, the breakdown of packages has no dependency on which
packages need which otehr packages. From a brief check (this is not
necssarily complete):
util/berk_db: needed by lib/kdb, lib/krb5
util/pty: needed by appl/bsd, appl/telnet
appl/telnet: needs util/pty and lib/krb5
On another note, the current method of calculating $(SRCTOP) and
$(BUILDTOP) is rather grotesque and may not work if we want to support
sub-packages. It is also responsible for much of the cruft in our
aclocal.m4 file. I think this is mostly a remnant of Imake-ism, but
admittedly it has some uses. Meanwhile, I am stripping out uses of
AC_PUSH_MAKEFILE because that mechanism is just one large gross hack.
Unfortunately many other macros depend on it.
---Tom
