ports/99818: Duplicated ports (ports/devel/otrs and ports/www/otrs)

[Nick Barkas]

On Thu, 06 Jul 2006 14:04:05 +0400
Sergey Skvortsov <skv at FreeBSD.org> wrote:
> Sorry for my inadvertence :(
> 
> Of course I ready to delete devel/otrs (I respect historical precedence).
> 
> But there are some notes:
> 
> 1. I suppose "devel" category is better than "www" because Bugzilla is
> placed inside "devel". Maybe it is not a good reason. Frankly, we need
> some new category like "www-apps" (like in Gentoo).

That makes sense. I noticed Bugzilla was in devel in the past, and
thought that was a little strange, too. At my job we use OTRS for
customer support, so being in devel didn't occur to me. If it is used
for bug tracking, or general issue tracking with developers, I suppose
it belongs in devel as much as Bugzilla does.

> 2. I'm not happy with "www/otrs"'s Makefile. First of all, this port was
> not even audited carefully before committing, for example:

Yes, I think your Makefile is better. I am very new to making ports
(this is only the third one I've worked on, and it is definitely the
most complex), so I made lots of mistakes.

> * to make OPTIONS really working you need wrap section with WITH/WITHOUT
> conditions with ".include <bsd.port.pre.mk>" and ".include
> <bsd.port.post.mk>"
> 
> * many modules inside Kernel/cpan-lib/ should be cleaned out, because
> they can conflict with installed from ports tree
> 
> * port's files was not patched to replace "/opt" with correct "${PREFIX}".
> 
> I suppose my port is a little bit cleaner but my option can not be very
> objective ;)

All three of these I agree with. I overlooked all of these problems
when making my port. Now I should not make the same mistakes next
time :)

> As a solution, I see two ways:
> 
> 1. I remove "www/otrs"
> 2. I can pass mainainership of "devel/otrs" to snb at threerings.net
> 3. Nick will merge all good features from "www/otrs"
> 
> or
> 
> 1. I remove "devel/otrs"
> 2. Nick merge all good features (if any) from "devel/otrs" into "www/otrs".

I think your devel/otrs port is probably what should be kept, and my
port can be removed. However, I have a couple of suggestions for your
port: First, it would be nice if the pkg-install script created the
otrs user with a fixed UID (I used 117, because it does not appear to
be taken by other ports in this list:
http://www.freebsd.org/doc/en_US.ISO8859-1/books/porters-handbook/dads-uid-and-gids.html).
The otrs user should be ok to just be put in the default www group
(80), I think, or should have a dedicated GID <1000 assigned, too. When
I installed devel/otrs on one of my systems to look at it, the UID and
GID assigned were over 10000, in a range that we use for LDAP user and
groups. Anyone using LDAP or NIS might have problems if an otrs
user/group ends up clashing with a LDAP/NIS user or group that is
created after the local otrs user with the same numeric UID or GID. My
pkg-install script also checks to make sure the reserved UID isn't in
use by a different user, and that its home directory is set correctly
if it has already been created. If a specific UID and/or GID is used, a
committer should probably update that list of reserved numbers, too :)

Second, the pkg-message in devel/otrs points out that the user needs to
configure Apache, but it does not mention more about the need to set up
a database, email, cron jobs, and possibly other things. It might be
nice to note these in the pkg-message, or direct the user to the
documentation installed in /usr/local/share/doc/otrs about the
additional post-install configuration needed.

> So, I wait for your suggestions/opinions.
> 
> Anyway, I ready to review any changes/patches from Nick to "otrs" port
> and commit them immediately.
> 
> -- 
> Sergey Skvortsov
> mailto: skv at FreeBSD.org
> 

Nick
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 191 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/freebsd-ports-bugs/attachments/20060706/e178bde7/attachment.sig>