Re: Bad rpmdb, where'd this come from?

[Jeff Johnson <jbj redhat com>]

On Sat, May 25, 2002 at 08:41:45PM -0700, Harry Putnam wrote:
> 
> Running 7.1 and seems to have been running ok for quite a while.
> Today running up2date -u as I do every so often I got a few updates.
> imap, imlib, and fetchmail and sharutils.
> 
> I realized I had no need for imap so just rpm -e'd those instead of
> installing the new ones.  When I did I got this message:
> 
>  rpm -e imap-devel
> 
>  rpmdb: Suspiciously high nelem of 4294967295 on page 0
>  error: db4 error(-30979) from db->verify: DB_VERIFY_BAD: Database verification failed
> 
> Then running rpm -UvhF *.rpm on what was left
> 
>  Preparing...                ########################################### [100%]
>     1:imlib                  ########################################### [ 20%]
>     2:fetchmail              ########################################### [ 40%]
>     3:imlib-cfgeditor        ########################################### [ 60%]
>     4:imlib-devel            ########################################### [ 80%]
>     5:sharutils              ########################################### [100%]
>  rpmdb: Suspiciously high nelem of 4294967295 on page 0
>  error: db4 error(-30979) from db->verify: DB_VERIFY_BAD: Database verification failed
> 
> The number after `high nelem of' in the first error may have been a
> different number.  I didn' actually keep that one, and used one that
> followed for example.
> 
> Anyone know what is happening here?

Scary message, ain't it?

Short answer:
	Message (and problem) is harmless, ignore.

Longer answer:
	The last deletion from a hash page did noty mark the page dirty,
	causing the reference count on the page to be one smaller than
	it should be. Later, when a verify is run, the -1 refcount is
	detected, and the condition is reported.

	There's a Berkeley DB patch to fix, dropped out when upgrading
	to db-4.0.14, reapplied since:

	* Wed Feb 13 2002 Jeff Johnson <jbj@redhat.com>
	- reapply Berkeley DB patch #4491.
	
What's a PITA, is that the condition is persistent in the database store,
so, if you've ever used a version of rpm w/o the fix, the message will
eventually be displayed, leading to all sorts of confusion about whether
the problem is actually fixed or not. (Hint: the problem is fixed).

73 de Jeff

-- 
Jeff Johnson	ARS N3NPQ
jbj@redhat.com (jbj@jbj.org)
Chapel Hill, NC