fp file locking behavior

[John Esak]

Bob,
I don't think the provisionary locking actually works that way. I mean, I
don't think it could cause any kind of deadlock or lockout condition.  I
could be wrong, of course, but I just don't think this situation could
happen.

What might be happening is a case where filePro is locking record zero
(which it does for various reasons... Maybe Ken can illucidate on this a
bit. It might bear some fruit for you.

John


> -----Original Message-----
> From: Bob Haussmann [mailto:haussma at nextdimension.net]
> Sent: Wednesday, May 23, 2007 7:36 AM
> To: john at valar.com; 'Filepro-List at Lists. Celestial. Com'
> Subject: RE: fp file locking behavior
>
>
>
> > It was JP Radley who first brought it to my attention that
> > automatic index
> > file stamps get changed just by someone entering into a file
> > in clerk. Yes,
> > it was long ago, but not more than 23 years or so... I found
> > a reference to
> > his notation about this back in some CompuServe stuff in
> > 1985.... and I
> > actually remember when he brought this up, thinking that it
> > was odd that the
> > stamps should change when no actual modification was made to
> > the file(s).
> > This behavior has never caused me a bit of hassle in all those years,
> > though.
>
> John--nice memory.  Now that you bring it up, I do remember that
> exchange.  I think it was after a major version update.  The question
> remains, why would fP need to issue a lock for writing the
> index when one is only browsing?  Interestingly, if PFREADONLY is
> set, the timestamp is not updated, so this behavior is likely by
> design.  Still, in very active files, wouldn't this cause
> deadlock/contention issues?  I think that is what we are seeing, and
> why it is so hard to track down.  It has to be two sessions hitting
> the same index at the exact same time, and one of those sessions
> encountering an index locked for writing.
>
> Bob Haussmann
> Tabor Children's Services, Inc.