Subject: Re: HEADS UP: Package removal candidates
To: Dieter Baron <dillo@danbala.ifoer.tuwien.ac.at>
From: Jim Wise <jwise@draga.com>
List: tech-pkg
Date: 12/06/2005 12:26:03
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
On Tue, 6 Dec 2005, Dieter Baron wrote:
>On Mon, Dec 05, 2005 at 06:49:24PM -0500, Jim Wise wrote:
>> On Tue, 6 Dec 2005, Thomas Klausner wrote:
>>
>> >Since Krister has been doing bulk builds for a long while now, I'm
>> >quite sure that he knows how to configure bulk builds correctly,
>> >so I'm convinced that something's missing or broken in the packages,
>> >not the bulk build setup.
>>
>> With due respect, the setting of `JAVA_HOME' is a basic requirement on
>> _all_ platforms on which java is used. It is as elementary as setting
>> PATH for unix command line work, or CFLAGS for C development.
>
> However, in pkgsrc, mk/java-vm.mk chooses which JVM is used and sets
>PKG_JAVA_HOME accordingly. devel/eclipse has the following line to
>make that setting visible to the package build:
> MAKE_ENV+= JAVA_HOME=${PKG_JAVA_HOME}
>Maybe that would help here also.
I see. Should we then add
MAKE_ENV+= CC=${PKG_CC}
to all package makefiles as well?
JAVA_HOME has been documented as a user settable variable for several
_years_ now, in pkgsrc/mk/{mk.conf,defaults/mk.conf}. If this is to
change, it is the responsibility of the _changer_ to:
a.) test old packages with the change
b.) make sure that standard variables are set by whatever mechanism
pkgsrc uses, just as our C Compiler choice support updates CC
and CFLAGS (and for the same reason: because these are standard
variables for a third party makefile to depend on).
c.) update the documentation
It sure looks like none of these have been done. If this is so, this is
hardly a flaw in packages which stopped working when these
infrastructure changes were made.
If JAVA_HOME is no longer to be a user settable variable, it is the
responsibility of our infrastructure to see that it is set when calling
a third party packages make infrastructure. It is not the
responsibility of package authors to guess which previously-documented
make variable has been removed this week...
- --
Jim Wise
jwise@draga.com
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2 (NetBSD)
iD8DBQFDlcmvhfG1+Pg8jmsRAsk0AKCikhcNekbvsJ0XudJCykeuZpmBvQCglDBp
1NdQAy5Sp4PsAbf8/EhFPQU=
=bJHH
-----END PGP SIGNATURE-----