Re: New RPM API - Documentation

[Jeff Johnson <jbj redhat com>]

On Fri, Apr 30, 2004 at 04:24:49PM -0400, James Olin Oden wrote:
> > It would seem to me that the non-priv side would need to do all the conflict 
> > resolution in order to insure that nothing stupid will happen (IF conflict THEN 
> > show_user_the_Problem) before invoking the root process.
> > 
> > Now, I suppose I could just pass off the list of new packages to install/upgrade 
> > and assume that the root process will reach the same conclusions as the non-root 
> > process. But we all know ASSUME makes an ASS of U and ME.
> >
> That seems reasonable, I am not sure but I think you can run rpmtsOrder()
> and rpmtsCheck() as a non-privileged user, such that you could do this
> before sending the data on what packages should go in the transaction to
> the priviliged process.
> 

Sure you can.

Both of rpmtsCheck() and rpmtsOrder() require read permission
on /var/lib/rpm, no other privs (like chroot(2)) needed.

> 
> P.S.  I am going to test the above...but I am experiencing some minor 
> difficulties.
> 

Well, you need write access to /var/lib/rpm/__db* to create locks
even if reading technically, but one can survive quite well w/o
the ability to create shared locks.

73 de Jeff

-- 
Jeff Johnson	ARS N3NPQ
jbj@redhat.com (jbj@jbj.org)
Chapel Hill, NC