Re: The value of direct inspection (was: Re: --initdb)

[Jeff Johnson <n3npq jbj gmail com>]

On Dec 2, 2006, at 10:54 PM, Tony Nelson wrote:

> At 6:19 PM -0500 12/2/06, Jeff Johnson wrote:
>
>
> > Meanwhile, --verifydb and "rm -f /var/lib/rpm/__db*" are well along
> > to being ripped out of rpm as well:
> >
> >     https://lists.dulug.duke.edu/pipermail/rpm-devel/2006-December/
> > 001923.html
>
> That is wise, just so long as the need for them is also removed.

And the need for --verifydb has largely been removed, patch at

    https://lists.dulug.duke.edu/pipermail/rpm-devel/2006-December/ 
001926.html

Hint: attempting a cron script to "automate" --verifydb is going to
create more problems than fixes imho, judging from my experience
running rpm -qa from cron. That cron script is going to be ripped as  
well,
now that /var/cache/hrmib (i.e. 0 length files named N-V-R.A with
install time stamped onto the file) is being reliably maintained as a
side effect of adding/removing headers from an rpmdb. Its gonna take
years to unlink -lrpmdb from net-snmp however, not the wisest patch
I ever wrote, sigh.

You really don't want to be messing with Berkeley DB, rpmlib, signals,
NPTL, "robust" posix mutexes, etc. Reproducers are almost non-existent,
and dynamic debugging across multiple processes is hard.

But have fun!

73 de Jeff