Re: Proposal - don't install .rpmnew unless changed

[Paul Howarth <paul city-fan org>]

Neal Becker wrote:
> Paul Howarth wrote:
>
>
> > Neal Becker wrote:
> >
> > > Paul Howarth wrote:
> > >
> > > > Neal Becker wrote:
> > > >
> > > > > Paul Howarth wrote:
> > > > >
> > > > > > Neal Becker wrote:
> > > > > >
> > > > > > > It would save admins a lot of time if we modify rpm so that it does
> > > > > > > not
> > > > > > > create a .rpmnew file if there is no change from the old file.  I
> > > > > > > would think this would be a simple modification.
> > > > > >
> > > > > > It already does this, doesn't it?
> > > > >
> > > > >
> > > > > I don't think so!  I keep syncing with develop every day, and most days
> > > > > I get a bunch of messages about "blah created as .rpmnew", and every day
> > > > > I run diff, and almost always get no output.
> > > >
> > > > Are you on an x86_64 box with lots of parallel-installed i386 packages?
> > >
> > >
> > > Yes, x86_64.  I have most parallel i386 packages that are standard on
> > > x86_64.  I did not install extra i386 packages.
> > >
> > > Today, for example, there were a bunch of messages about upgrade to
> > > kdelibs-3.5.0-5.  It does happen that there are both x86_64 and i386
> > > versions of this, do you think this is the explanation?  In any case, any
> > > chance to fix it?
> >
> > I think it's the same issue (multiple packages owning the same config
> > file) as for /etc/vimrc, except in this case it's different-arch
> > packages instead of different-name packages.
> >
> > I don't know what the *right* policy for these cases should be really.
>
>
> Isn't it "right" to not create .rpmnew and report it to the user if there is
> no diff to the current version?

What *is* the current version though, if there are multiple packages 
owning the same file?

I don't know enough about rpm internals to say why this is a problem.

Paul.