http://www.slony.info/bugzilla/show_bug.cgi?id=273

Christopher Browne <cbbrowne at ca.afilias.info> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|slony1-bugs at lists.slony.inf |janwieck at yahoo.com
                   |o                           |
             Status|NEW                         |ASSIGNED

--- Comment #2 from Christopher Browne <cbbrowne at ca.afilias.info> 2012-11-22 14:35:50 PST ---
The patch seems a bit counterintuitive, in that it's looking for absence of
node.  That's not the same as saying it's wrong; just that it's going to take
more careful reading to validate things.  And as this seems to be deep in the
multi-threaded parts, I think I'm inclined to trust Jan's opinions more than
mine.

I would think it more ideal if there were two patches:

a) The one that fixes things (which I imagine you already have there), and

b) Another which makes it as easy as possible to induce the conditions that
tickle the problem in the DisOrder tests.

This might, for instance, include pushing some SYNC events at node #1, and
adding/dropping a subscription of node #3 against node 5 a few times, perhaps
with the merge.

If we already have a test doing that, and I think we do in MergeSet.js, then
perhaps there is already something adequate.  Or perhaps if we add an
iteration, that can encourage the error condition to emerge with greater
probability.

It's particularly nice if we can have a good test that:
a) Fails nearly 100% of the time without the fix, and
b) Fails nearly 0% of the time with the fix.

-- 
Configure bugmail: http://www.slony.info/bugzilla/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.
You are the assignee for the bug.