Re: [Sks-devel] DB_INIT_LOCK problems?

[Yaron Minsky]

On Sat, 31 Jul 2004 12:30:47 -0400, Jason Harris <address@hidden> wrote:
> On Fri, Jul 30, 2004 at 10:10:40PM -0400, Yaron Minsky wrote:
> > On Fri, 30 Jul 2004 19:15:45 -0400, Jason Harris <address@hidden> wrote:
>
> No, one single-threaded process writing a single record which updates
> multiple pages will deadlock its own self, due to the internal design
> of BDB's locking subsystem, even if it is the only process using a
> database.
> 

Freaky.  I guess that means you always need to run the deadlock
detector whenever you're using locking.  And, further, that you need
to tolerate interrupts from the deadlock system.

Sigh.  Seems like a royal pain to get this all working properly.