On Thu, 28 Oct 1999, Steve Philp wrote:
> Axalon Bloodstone wrote:
> >
> > On Wed, 27 Oct 1999, Steve Philp wrote:
> >
> > ^ is that a type-o ? i have 2mdk and 5mdk localy (i'd really hate to think
> > i'm that outdated, but it is posible)
>
> Here's the header from rpm -qi initscripts, the package comes from the
> MacMillan 6.5 Deluxe.
>
> Name : initscripts Relocations: (not
> relocateable)
> Version : 4.23 Vendor: MandrakeSoft
> Release : 33mdk Build Date: Tue Sep 7 15:15:29 1999
=X wow i've had the same "basic" install for a whole month, truely
amazing..
> > > I was messing with /etc/rc.d/init.d/network tonight trying to help out
> > > another user to get aliased interfaces to work correctly. I had hacked
> > > the file to bits without creating a backup (shame on me, I know).
> > >
> > > I used 'rpm -e --nodeps' to remove the initscripts package thinking I
> > > could easily reinstall it from the CD. How wrong I was.
> > >
> > > I used 'rpm -i --nodeps --force' to install the package but
> > > /etc/rc.d/init.d/network didn't reappear. Actually _NONE_ of the files
> > > contained in the package got installed (rc.local, for instance, didn't
> > > get replaced -- my changes were still at the bottom of the file).
> >
> > I didn't need to use --nodeps or --force, the latest versions i know do
> > not require linux_logo anymore so you shouldn't ever need tell it nodeps.
>
> I had to use --nodeps to remove the package because the kernel depended
> on it. I had to use the --force on the installation because RPM
> believed it was still installed. linux_logo was removed along time ago
> -- until the wraparound on the processors line is fixed, I refuse to use
> it.
Yep knew the --nodeps for rpm -e, lots depend on initscripts
(see below)
linux_logo, wraping one Celeron (and SMP) has been fixed.. (didn't wrap
here, of course it always gets replaced by Welcome2l on my boxes)
> > > The --nodeps in both of the above lines are necessary because RPM
> > > (rightly) believes that the package is necessary. The --force is
> > > necessary upon reinstallation because RPM (wrongly) still believes the
> > > package is installed.
> >
> > stupid rpm bug. it does not like the
> >
> > [ -f /var/lock/TMP_1ST ] && rm -f /var/lock/TMP_1ST
> >
> > for the %postuninstall (bash2 issue if i remeber right)
>
> It doesn't sound like the same sort of problem that's affecting the
> portmap package then. In the initscripts case, the package no longer
> shows in the package listing. In the portmap case, the package
> continues to be shown despite not being installed any longer.
Yes if it doesn't sow up thats a totaly different problem, however if it
doesn't show why still --force, does it say it is installed but not show
on rpm -q or rpm -qa?
If either uninstall script fails, rpm says the whole thing fails (why it
doesn't always honor this with -i but does with -e, i am quite curious)
Oh yeah, for that pesky portmap you still have dangleing in the db,
rpm -e --noscripts --justdb portmap
should work
> > > Thinking it might have been a bad package, I installed the source
> > > package and rebuilt it with 'rpm -bb' then reinstalled it. Still no
> > > change. RPM thinks it installed the package!
> > >
> > > Finally, I ended up just going into the BUILD/ directory and hand
> > > copying the files I needed back into their directories.
> >
> > rpm -bi --short-circuit initscripts.spec
> > cp -r /var/tmp/initscripts-something/* / # <- this isn't right but you get
> > the idea
>
> I just did a 'make install' in the BUILD/ directory. Seemed like the
> direct path to me! :)
Is safe if the pkg uses DESTDIR, otherwise it's posible the specfile will
modify Makefile's, also you get things we leave out and miss some we might
add (even the above doesn't account for the former)
> > > Is something horribly broken with the RPM database that it's not getting
> > > updated when packages are removed? I previously reported a bug with
> > > portmap in which it's entry didn't get removed from the database when
> > > the package was removed -- any progress on that one?
> > >
> > > So, the question is -- How did initscripts files get installed in the
> > > first place if they won't install now? How do I GET that package to
> > > reinstall correctly?? I can provide 'rpm -Uvvvvvh --nodeps --force'
> > > output if necessary...
> >
> > (touch /var/lock/TMP_1ST \
> > rpm -e initscripts --nodeps && \
> > rpm -i initscripts-4.42-3mdk.i586.rpm ) || ( \
> > rpm -i --replacefiles --replacepkgs initscripts-4.42-3mdk.i586.rpm )
> >
> > Something like that, you probably should never ever rpm -e initscripts
> > just imagine had your power gone off :)
>
> No biggie. Installs don't scare me.
>
> > Q: How did initscripts files get installed in the first place if they won't
>install now?
> > A: the initial packages (base section, from comps) are installed via cpio
> > not via rpm
> >
> > Not sure about subsequent rpms..
>
> Makes sense I suppose. I'm still not sure about initscripts being
> correct yet. I'll take another look through the spec file to see if
> anything obvious jumps out at me.
>
>
--
MandrakeSoft http://www.mandrakesoft.com/
--Axalon
