RPM broken

[James Olin Oden]

On Sat, 3 Apr 2004, Brian Bober wrote:

> > Date: Sat, 3 Apr 2004 18:34:59 +0100 Martin Alderson  wrote:
> > Subject: RPM broken
> > 
> > I seem to remembering having to clear a cache or lock file, or maybe 
> > even rebuilding the RPM database.
> > 
> 
> On former distributions (as root), this worked, but it won't now:
> rm -f /var/lib/rpm/__db*; rpm --rebuilddb
> 
> Because the __db files aren't located there...
> [boberb at localhost boberb]$ slocate __db
> /usr/lib/rpmdb/i386-redhat-linux/redhat/__db.002
> /usr/lib/rpmdb/i386-redhat-linux/redhat/__db.003
> /usr/lib/rpmdb/i386-redhat-linux/redhat/__db.001
>
Those are from the rpmdb-redhat (or perhaps fedora) package.   Yes, the 
rpm database for your system is still located in /var/lib/rpm, and yes if 
you have a stale lock on the rpm database and delete the __db* files 
(which are not themselves lock files...ask Jeff Johnson to explain if you 
really want to know) the stale locks will go away (really held futex's).
Additionally, there is the new dynamic with the transaction lock file, 
that Paul mentioned.

> And considering they are in a different place, its questionable whether the
> same things holds true. I wouldn't mess with them.
>
You misinterpreted what you read.
 
> 
> I also suggest updating rpm the first thing you do when installing fedora 2
> test2 (as root):
> 
> yum install rpm
> 
>
Not a bad idea.  That said, if its a db lock, though, won't fix the 
problem, because
its either buggy code using librpm, or something using librpm got killed 
with SIGKILL.  If its the other lockup I am seeing with scriptlets, 
updating won't likely help either.  If its something to do with 
transaction lock...well, I don't know you would have to ask Paul Nasrat 
(hope I didn't mispell your last name) or Jeff Johnson who are working on 
that.

Cheers...james
>