db-torque-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Thomas Fischer <fisc...@seitenbau.net>
Subject Re: TRQS321: Criteria Serialization
Date Thu, 10 Nov 2005 08:52:06 GMT


I did not see you already provided a patch in scarab. I will look at it
today evening. If noone objects, I will include it into the release.


Thomas Fischer <tfischer@apache.org> schrieb am 10.11.2005 09:19:06:

> On Thu, 10 Nov 2005, Scott Eade wrote:
> > Thomas Fischer wrote:
> >
> >> I took a quick glance at the issue, and when I uncommented the test
> >> you checked in and resolved an issue of visibility (use
> >> "Criteria.getNewCriterion()" instead of "new Criteria.Criterion"), the
> >> case did not produce an error.
> >>
> >>
> > This is interesting.  What JVM are you using?  I am using J2SE 1.4.2_06
> > from Sun for Win32.
> I could reproduce the error at a second machine. It uses 1.4.2-05 for
> linux.
> >
> >> Then I looked what the test case does, and it uses
> >> to test whether the two criteria are equal. This is not enough, since
> >> Criteria.equals does not compare aliases and asColumns. There are two
> >> possibilities to remedy this:
> >> - The first is to change criteria.equals, which I did not want to do
> >> between release candidates, but if somebody else is of another opinion
> >> here,  we need to discuss it.
> >>
> >>
> > If equals() is performing its function correctly then this is IMO a bug
> > and most bugs warrant fixing - even between release candidates (though
> > preferably not between the last rc and the final release).
> ok, now that I can reproduce the error, I see that Criteria.equals() is
> not the problem (it is not strict enough in my opinion, but this has
> nothing to do with the serialisation problem.)
> > ...
> > As I said earlier, this has been broken (at least for me) for a while,
> > wouldn't hold off 3.2 for this issue as nobody spoke up earlier or felt
> > that it was worth investing their own time to fix (me included).
> If we want to solve this problem, we can as easily fix it before the
> release as after. So let's do it now, if we can.
> > According to Bloch (Effective Java Item 54) the default serialized form
> > of an inner class is ill-defined.  This could lead to varying behaviour
> > between different JVMs.  The answer may be to add custom serialization
> > code to Criterion.
> I could not get hold of Bloch's book in short time, but according to what

> I have found on the internet, the problem with inner classes only occurs
> for non-static inner classes. So I thought maybe one can fix this by
> making Criterion a static inner class (diff appended), but this does not
> solve the problem (same behaviour as before).
> I am puzzled at the moment and need to dissect the problem further to
> understand it. Not sure whether it will happen before the week-end,
>     Thomas[Anhang "patch.txt" gelöscht von Thomas
> Fischer/kn/seitenbau]
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: torque-dev-unsubscribe@db.apache.org
> For additional commands, e-mail: torque-dev-help@db.apache.org

To unsubscribe, e-mail: torque-dev-unsubscribe@db.apache.org
For additional commands, e-mail: torque-dev-help@db.apache.org

View raw message