Re: building a package that can not be upgraded

Stuart D Gathman <stuart@xxxxxxxx> · Fri, 06 Jul 2012 13:49:19 -0400



    On 07/05/2012 05:09 PM, Valery Reznic expounded in part:
    
      
        You can write pre-install script that check number of
            installed package instances ($1, iirc) and if
        package already installed, then exit with error
            message and non-zero status.
        

        But why do you want to do such a thing?
        

    Like /dev/null and /dev/full, such boundary conditions often have
    uses.  For instance, how about an RPM package that can't be
    *installed*?  Why would you want such an apparently useless thing? 
    Well, right off, I could see it being useful in some repository to
    replace a broken package - effectively also preventing anything
    depending on it from being installed with a clean error.  (Although
    simply removing the broken package from yum repo results in a
    reasonable error message.)  

    
    In fact, the use case for "can't be upgraded" in this instance is a
    broken package.  Eventually, the package will be fixed, and *then*
    you will want to upgrade it.  So maybe you want to create a new
    release of the current (broken) version that can't be installed. 
    That will prevent updates with yum until you get it fixed.

  
_______________________________________________
Rpm-list mailing list
Rpm-list@xxxxxxxxxxxxx
http://lists.rpm.org/mailman/listinfo/rpm-list