Oplock bug?

[Jeremy Allison]

On Mon, Jul 18, 2005 at 11:42:11AM -0700, Jeremy Allison wrote:
> 
> Sorry Volker, looks like this is a bug you introduced when simplifying the
> code.... :-(.

Ok, looking closely at this - you ran into one of my paranoia
defenses when I was creating the original code. There is no
real reason to keep an independent count of the number of
outstanding oplocks - only paranoia to ensure the numbers
match. We could just ignore that test and debug statement
and always call oplock_break(), which would return True
on not finding the oplock - then return the message 
and the logic would look much simpler.

I apologise for how obscure this logic was - but you have
to remember I was creating this code whilst working out
the correct logic, and so keeping a separate count allowed
be to put independent debug checks in place (like the one
you ran into) that confirm the underlying logic in two
different places.

It should be much better when your tdb messaging replacement
is in place for 3.0.21 and I'm looking forward to that.

Sorry,

	Jeremy.