Hi Clint, Clint Adams schrieb am Mon 12. Jan, 14:33 (+0000): > On Mon, Jan 12, 2009 at 01:59:08PM +0100, Jörg Sommer wrote: > > I've enabled core dumps and it looks like a problem in libdb. I was to > > rashly with the guess of an ABI problem (that was what I've meant with > > linking problem). > > > #0 __memp_bh_priority (bhp=0x11b9d92b) at ../dist/../mp/mp_mvcc.c:30 > > Thanks; there may be a relevant fix in 4.6.21-13. Could you check > to see if that version solves your problem?
As I can't reproduce the problem, so I can't surely if it would be gone,
but I try the upgrade. But the changelog doesn't sound like anything
around this was changed.
[New process 15522]
#0 __memp_bh_priority (bhp=0x11b9d92b) at ../dist/../mp/mp_mvcc.c:30
warning: Source file is more recent than executable.
30 while (SH_CHAIN_HASNEXT(bhp, vc))
(gdb) l
25 __memp_bh_priority(bhp)
26 BH *bhp;
27 {
28 u_int32_t priority;
29
30 while (SH_CHAIN_HASNEXT(bhp, vc))
31 bhp = SH_CHAIN_NEXT(bhp, vc, __bh);
32
33 priority = bhp->priority;
34
I think the chain is broken and the last pointer points to an invalid
region. But I didn't see how to verify it. The code in the upper frame is
this:
(gdb) l
734 bhp->pgno = *pgnoaddr;
735 bhp->mf_offset = mf_offset;
736 SH_TAILQ_INSERT_TAIL(&hp->hash_bucket, bhp, hq);
737 SH_CHAIN_INIT(bhp, vc);
738
739 hp->hash_priority =
740 BH_PRIORITY(SH_TAILQ_FIRSTP(&hp->hash_bucket,
__bh));
741
742 /* We created a new page, it starts dirty. */
743 if (extending) {
Do you have an idea?
Bye, Jörg.
--
chinesiches Sprichwort:
Wer fragt, ist ein Narr für fünf Minuten.
Wer nicht fragt, ist ein Narr fürs ganze Leben.
signature.asc
Description: Digital signature http://en.wikipedia.org/wiki/OpenPGP
