[TriLUG] rpm database question

[Adrian Likins]

On Tue, Nov 20, 2001 at 04:50:36PM -0500, Lisa Lorenzin wrote:
> 
> for folks running redhat, how do you reconcile wanting to compile your own
> versions of things with wanting to keep the rpm database up to date?
> 
> for example, i'm seriously considering installing the latest version of
> sendmail on my new RH7.2 box.  there's not an rpm for it yet, and even if
> there was, i'd like to make some customizations to it (nothing fancy -
> just cosmetic).  at that point, what my rpm database knows about will be
> out of sync with what my box has, which will screw up my automatic updates
> via rhup and confuse me mightily when i go to do something else with
> sendmail in 6 months, having totally forgotten i did this.  (okay, with my
> memory, make that 2 months.)
>
	for me
		home box: 
		   whatever, dont care... lots of rpms, lots of
		   stuff built from cvs, lots of tarballs. Sometimes it's
		   pain getting new stuff to build or upgrade

		work box:
		    rpms only for anything work related. occasionally
                    cvs or tarballs for unimportant things (a gimp
		    plugin, etc). 		   

		Production Boxes:
		   _Nothing_ goes on the box that isnt an RPM. 
 
	ie, anything important I install as an RPM. 

> so, is there any way around this?  do i have to just choose between
> compiling my own versions and having rpm reflect reality, or can i have my
> cake and eat it too?
>
	There is support for creating virtual packages of stuff installed manually,
but it's probabaly easier just to build a rpm of it. Probabaly about the same
amount of work involved, but you get the benefits of being able to sign the
package, have the md5sum of the files, a fairly reproduceable build, and a package
you can easily reinstall if need be.


Adrian