directory-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Selcuk AYA <ayasel...@gmail.com>
Subject Re: Moving txns boundaries into the DefaultCoreSession class : heads-up
Date Fri, 23 Dec 2011 06:56:19 GMT
On Thu, Dec 22, 2011 at 11:36 AM, Emmanuel Lecharny <elecharny@gmail.com> wrote:
> Hi guys,
>
> so, as suggested by Selcuk yesterday, I'm moving the txns handing into the
> DefaultCoreSession class. That impacts the LDAP handlers, as I had to remove
> the beginTxn/commitTxns/abortTxns from those classes, plus move the logic
> inside the DefaultCoreSession class.
>
> I have started doing so with Add/Delete/Exists/Lookup code, and so far, it
> seems to be ok. Although until all the other operations have been procesed
> (ie lookup, search, modify, etc) I will have more errors than previoulsy,
> just because if all the operations aren't protected by txns, there are side
> effects.
>
> For instance, I tried to move the add and delete txns boundaries only, but I
> got some tests failure because those tests were doing some lookup() and
> exists() which weren't protected by txns. Funny enough, the tests were
> passing when I was debugging them, due to some delay I introduced simply by
> being slow when stepping through the code, but were wailing when I ran the
> test without break point.

There is a background thread that flushes pending txns in the txn log
to the underling partitions. That is probably the reason you have
different test results when stepping through code. You can disable
that thread to make things more predictable when debugging.

>
> Anyway, I will continue today and will provide some heads-up.
>
> --
> Regards,
> Cordialement,
> Emmanuel L├ęcharny
> www.iktek.com
>

Mime
View raw message