discuss-gnustep
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: $GNUSTEP_DEFAULTS_DIR not set/exported in GNUstep.sh


From: Rogelio M . Serrano Jr .
Subject: Re: $GNUSTEP_DEFAULTS_DIR not set/exported in GNUstep.sh
Date: Tue, 03 Aug 2004 00:38:11 +0800

On 2004-08-03 00:23:24 +0800 Armando Di Cianno <fafhrd@gentoo.org> wrote:

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 2004-08-01 23:49:13 -0400 Sheldon Gill <sheldon@westnet.net.au>
wrote:
It is actually very easy to change with my path utilities patch. There's a simple DEFAULTS_DIR declaration which you can change and will then apply for all users on the machine. So this will be fixed RSN without needing any mods to -make.

Which patch is this?  Is it available somewhere, or something that was
already committed?

I've been trying and trying to find a solution for (what I'm now
convinced is) this mess, but it's really bizarre, and kludgy, that we
use both an evironment generating script (GNUstep.sh) and a file
(GNUsteprc) for some of the same setup ideas.

I looked through the sources, and the use of GNUsteprc still doesn't
elegantly solve my problems with building applications in a sandbox.
I'm curious why we don't just choose one or the other?

... and I still say user_home.c and NSUser.m should pay attention
to$HOME ... I guess that's my problem with having this both ways: I
can't just "pass in" through the environment GNUSTEP_DEFAULTS_ROOT,
'cuase it seems to only be settable from GNUsteprc.  And since there's
a GNUSTEP_USER_ROOT settable somewhere, using getpwnam() on *NIX's,
otherwise, sort of doesn't make sense -- why have it definable through
an rc-file, but then assume the entry from /etc/passwd file for
everything else?  If it's coming from /etc/passwd, then it should've
made it into $HOME .... which of course _is supposed to be_ definable
for progams that write user setting into /home/whatever; think chroot
env's, daemons running as 'nobody' w/o a home, etc, etc. We should
make these settings either dynamic (changable everytime an program is
called) or static (changable via rc file) but not both.  Using an rc
file doesn't really make things dynamic.

How about we just decide on what environment/setting varibles should
exist, and then stick with them across the board, regardless of use of
environment or rc file?

If we have to stick to both methods, let's assume this:
GNUsteprc --(overridden by)--> environment
... and let's do nothing to the contrary.

Sorry if I seem a bit tense on this subject -- it's really rather
irratating, though.  Otherwise, GNUstep has me in all smiles. :-)

__armando
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)
Comment: Using the GPG bundle for GNUMail.app

iD8DBQFBDmp7wgiTPLI9xhcRArcxAJ4tNe2Q9XSvkFXd/4fk06CAHZvd/wCgmnXJ
lD9Eag02ehTRJX6jskR1HDg=
=x3rp
-----END PGP SIGNATURE-----



_______________________________________________
Discuss-gnustep mailing list
Discuss-gnustep@gnu.org
http://lists.gnu.org/mailman/listinfo/discuss-gnustep


Why dont we use compile time defaults for each platform GNUstep runs on and get rid of runtime configs? For linux why not use /etc/GNUstep/ or $GNUSTEP_LOCAL_ROOT/Library/?. User specific defaults are in in /home/$username/Library anyway. All domains are supposed to have the same heirarchy right? Then all GNUstep programs can use NSPathUtilities to find the directories they need. GNUstep base is supposed to hide the host system intricacies from Apps.





reply via email to

[Prev in Thread] Current Thread [Next in Thread]