On Aug 29, 2006, at 1:33 PM, Christian Goetze wrote:
The problem surfaced in March, and is part of the justification
for having all files depend
on their parent directory to avoid relocating orphan directory paths.
You can try reporting the bug against RHEL if you want the
problem fixed, that might work.
I may do that, although it seems at least for now, I'm relatively
happy..
Good.
Otherwise, I'd suggest using --root (which will require root
access because of chroot(2)) to
install rather than trying to use --relocate as you are doing,
that is very likely to Just Work.
I don't really want to do that, because I want developers to be
able to install the rpms as non-root within their source tree
(which is one of the reasons I use relocatable rpms).
Now one bug I really do wish they'd fix soon is the one where /var/
lock/rpm/transaction is hardwired. That lock file belongs in the
database dir, IMHO - what's the point otherwise?
Yep. That file in your environment probably gets in the way a lot.
The fcntl lock on that file serializes rpm installs, and was
requested to "stabilize"
rpm.
It was easier to add the lock than to try to understand what
"stabilize" meant.
The fcntl lock is now normally not used in (iirc) rpm-4.4.6 or so,
although one
who desires to "stabilize" can still do so if desired.
I know of no definite reason then or now why the additional fcntl
lock is needed.
OTOH, I can't pre-compute or test every code path through rpm.
73 de Jeff
_______________________________________________
Rpm-list mailing list
Rpm-list@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/rpm-list
