[Netatalk-devel] Re: Some configure patches and request for comments

[jeff b <je...@un...>]

Alain RICHARD wrote:
> no, in fact it worked only when replacing a variable with a 
> constant, but not when replacing a variable with another one. For 
> example, in rc.atalk.redhat.in, you may use @prefix@ and 
> @CONFIG_DIR@ that get replaced with /usr and /etc/atalk, but you 
> can't use @sbindir@ and @sysconfdir@ because they are replaced with 
> $prefix/sbin and $prefix/etc. The problem here is that $prefix is 
> not defined at the rc.atalk.redhat execute time. This was the main 
> reason this script has hardcoded some paths as 
> @prefix@/sbin/atalkd, making it incompatible with the various 
> configure normal functionality. Also all the man page and comments 
> in afpd.conf file where refering to :SBINDIR: variable for example.

I was aware of the substitution problem, which is why those paths are
hardcoded in there. If you have a better way of doing it using the
autoconf system, I'd like to see it.

> > My last tarball RFC releases were made without the original 
> > Makefile system. I haven't removed them yet because the 
> > ./configure system wasn't always working, but I suppose we can do > > that now.
> 
> I think so. Also we should make the dist target working and a 
> VERSION systems to differenciate between various tarball (what 
> about netatalk-1.5-20000918 ?).

Fine with me. I just have been too lazy to implement it in the build
script that I have. I'll do that as soon as I get to it.

> > Definately agreed. A unified Changelog would make *everything* so 
> > much easier.
> >
> > In terms of documentation, I am getting ready to commit a 
> > man/man5 directory (with ./configure and Makefile.am changes) 
> > with some man pages I have been writing for configuration, since 
> > netatalk documentation is fairly sparse.
> 
> What do you think about setting up an old directory with the old
> README and ChangeLog and starting a very simple README, INSTALL and 
> a new ChangeLog for the developement version 1.5 ?

That is a disgustingly good idea. The current documentation is a little
two spread out, and isn't too consistant. Having the old UMich changelog
in there, for example, isn't that useful for casual viewing.

> > >  5) Unify the default installation tree and make it compliant 
> > > with gnu standard. Mainly there should be no executables and 
> > > only config files in the config directory ( /etc/atalk 
> > > directory), so we should move the uams directory to an other 
> > > place (in the $libexecdir directory I think) and perhaps nls 
> > >files to $datadir.
> >
> > Perhaps uams to $libdir/atalk/uams/, nls to $libdir/atalk/nls/ ?
> 
> Here there are two cases : either the package is compiled to be
> installed in an autonomous directory, in that case I think we 
> should use the gnu standard :
> 
> ./configure --prefix=/opt/netatalk
> 
> /opt/netatalk/etc/afpd.conf
> /opt/netatalk/bin/nbplkup
> /opt/netatalk/sbin/afpd
> /opt/netatalk/lib/libatalk.a
> /opt/netatalk/libexec/uams/uams_guest.so
> /opt/netatalk/share/nls/maccode.iso8859-1
> /opt/netatalk/man/man8/afpd.8
> 
> The second case is when installing in a rpm based filesystem (for
> example a redhat linux). In that case, I think the spec file should
> configure as this :
> 
> ./configure --prefix=/usr --libexecdir=/usr/lib/atalk
> --sysconfdir=/etc/atalk --datadir=/usr/share/atalk
> 
> /etc/atalk/afpd.conf
> /usr/bin/nbplkup
> /usr/sbin/afpd
> /usr/lib/libatalk.a
> /usr/lib/atalk/uams/uams_guest.so
> /usr/share/atalk/nls/maccode.iso8859-1
> /usr/man/man8/afpd.8
> 
> We may also makes things simpler by setting the second file layout > the default.

Agreed. I don't have any problems changing my spec file to do that.

jeff