[CentOS] Centos Gotcha: YUM Groupinstall

Wed Aug 17 02:20:50 UTC 2011
David <david at daku.org>

At 07:13 PM 8/16/2011, you wrote:
>At Tue, 16 Aug 2011 18:32:43 -0700 CentOS mailing list 
><centos at centos.org> wrote:
>
> >
> > Folks
> >
> > I have encountered a situation with YUM that isn't what I expected.
> >
> > Let's suppose I want to install a group, call it G.  My first
> > question would be -- is the group already installed.  Realize all of
> > this is scripted.
> >
> > So, I use
> >    yum groupinfo
> > and I see the list of installed groups, and those not yet installed.
> >
> > If group G is in the list of installed groups, one would think that
> > there's no point in issuing a "groupinstall".
> >
> > But, to my surprise, a group can be listed as "installed", only to
> > find that a groupinstall will actually do a lot of installs.
> >
> > So, I had to adapt my script to perform a "groupinfo G", parse the
> > modules, and individually test them to see if they need installation.
> >
> > I do not understand this behavior -- group G is shown as "installed",
> > but it really isn't.
>
>I think what is happening is that some sub-set of group G, probably to
>satisfy various dependcies for rpms in group A, B and F (say).  Elements
>of Group G have been installed, but not ALL of group G has been
>installed, and thus group G is *partitically* installed.  In a sense
>group G is neither uninstalled nor fully installed.  This 'group'
>business is not 'atomic' as are individual rpms are 'atomic': a single
>rpm package is either installed or not installed -- you cannot properly
>install 'part' of an rpm, but you can install 'part' of a group [of
>rpms].
>
> >

Robert

Yes, I sort of came to the same conclusion, but couldn't express it 
as elegantly as you did.  Luckily, about 30 lines of  Perl took care 
of it, by converting the module list from groupinfo into individual 
items which I could match against the "yum list installed"