Re: [fink-core] passwd

[Hanspeter N. <han...@sn...>]

On 7/20/11 2:30 PM, Daniel Macks wrote:
> On Wed, 20 Jul 2011 09:10:58 -0700, "David R. Morrison"  wrote:
> Dear fink core,
>>
>> I have commited the distributed password package (thanks nieder!) to
>> 10.7.  There is a package passwd-core, and then lots of splitoffs
>> like passwd-fink-bld and passwd-games which take care of installing
>> specific users and groups.
>>
>> My idea for a strategy about this is to revise our passwd package in
>> 10.5 and 10.6 with a large "Provides" line including all of the
>> splitoffs of passwd-core.  That way, a maintainer can revise
>> "Depends: passwd" to "Depends: passwd-games" and get something which
>> works the same in all distributions.
>
> Sounds like a good plan. Definitely nicer to be able to say
> "dep:passwd-foo" than "dep:passwd (>= some magical version when foo was
> first added)" and maintain something that works in both the 10.[56] and
> 10.7 trees.

One thing I noticed a while back when I first created passwd-core was 
that packages were very non-standard in how they (Build)Depended on 
passwd.  This list is from a post of mine a year ago (exact contents may 
have changed since):

Depends
-------
alienblaster
angband
avahi*
cmine
dovecot
emacs*
gdm*
gnome-libs
icecast2
leafnode
moria
nethack
roguedarwin
rt-pm
tome
tornado
trackballs
warp
xbomb
xinvaders

BuildDepends
------------
xkobo

Depends and BuildDepends
------------------------
distcc
distcc-gtk
opennms
tomcat*
postgresql8*

BD in main package, D in splitoff
---------------------------------
dbus
canna

Expect some dependency on passwd, but don't have any declared
-------------------------------------------------------------
icecast (icecast2 does have Depends:passwd, maybe this version doesn't 
need it)
amqplib-py*

Other
-----
libhal1-shlibs (has DescPackaging FIXME! about passwd)

Concurrent D: and BD: seems useless.  BD only also doesn't make sense 
since most packages actually need the user they created to run. 
Depends: seems the best solution, at least while we have no method of 
actually tracking whether a user:group is actually created.

Hanspeter

-- 
Hanspeter Niederstrasser
hanspeter AT snaggledworks DOT com