rpm database corruption (again)

[Jeff Johnson]

On Mar 8, 2007, at 7:23 AM, Neal Becker wrote:

> On Thursday 08 March 2007, Jeff Johnson wrote:
>> On Mar 5, 2007, at 9:28 AM, Neal Becker wrote:
>>> On Monday 05 March 2007, Pascal Bleser wrote:
>>>> Neal Becker wrote:
>>>>> I'd almost given up on smart.  Running on fedora fc6.  I'd do
>>>>> smart upgrade --update
>>>>>
>>>>> And almost daily I'd have a broken rpm database.  Fixing it is
>>>>> painful.
>>
>> "... broken ... painful"
>>
>> Without describing "broken" in more detail, it's impossible to  
>> attempt
>> to diagnose your problem.
>>
>> 73 de Jeff
>
> smart update hangs at ~50% loading cache.  rm /var/lib/rpm/__*, then
> rpm --rebuilddb fixes it.
>

You likley don''t need --rebuilddb.

This is likely the kernel-2.6.19 (and FC6 2.6.18 backport) mmap(2)  
problem,
bugzilla is full of reports.

Verifying hung on futex with strace/gdb, or (almost equivalently) doing
     cd /var/lib/rpm
     /usr/lib/rpm/rpmdb_stat -CA
would confirm.

FYI, stale lock detection and removal is implemented in rpm-4.4.8-0.4
and the backport to 4.4.2 is quite feasible. I'd be happier to do the
backport then have to listen to guesses regarding "corruption"
causes for the next year or two.

But that's Fedora''s, not rpm's, problem. Shrug.

hth

73 de Jeff