openjpa-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "robert burrell donkin" <robertburrelldon...@gmail.com>
Subject Re: PCRegistry ClassLoader memory leak
Date Tue, 17 Jul 2007 17:12:11 GMT
On 7/17/07, Kevan Miller <kevan.miller@gmail.com> wrote:
>
> On Jul 17, 2007, at 12:15 AM, Craig L Russell wrote:
>
> >
> > On Jul 16, 2007, at 7:48 PM, Kevan Miller wrote:
> >
> >>
> >> On Jul 16, 2007, at 10:26 PM, Pinaki Poddar wrote:
> >>
> >>> Just to clarify:
> >>> I meant it should be the loader that loaded Person.class where
> >>> Person is
> >>> the persistence capable class, *not*
> >>> org.apache.openjpa.enhance.PersistenceCapable.class.getClassLoader
> >>> ().
> >>>
> >>> But testing in multi-classloader environment is required to
> >>> validate any
> >>> changes of this nature.
> >>
> >> Yes, that's how I interpreted your statement. My point is that if
> >> SSN is a field of Person, the ClassLoader of SSN is not
> >> necessarily the ClassLoader for Person. I'd be concerned that the
> >> Person ClassLoader can't load SSN. In that case, your technique
> >> wouldn't work...
> >
> > Just to clarify, if SSN is the type of a field of Person, the
> > ClassLoader of Person.class must be able to load SSN (either itself
> > or via a parent ClassLoader) or you will have a linkage error while
> > loading Person.
>
> Craig,
> You are correct. The declared type must be loadable. I was thinking
> of an SSNImpl type which need not be. However, that's not really
> relevant to the problem at hand...
>
> Pinaki,
> I'm afraid that I may have improperly narrowed your objectives by
> singling out fieldTypes. PCRegistry$Meta.pc and pcSuper could also
> keep Classes/ClassLoaders alive...

if there are several places where this may happen then perhaps  try a
variant on http://svn.apache.org/repos/asf/jakarta/commons/proper/logging/trunk/src/java/org/apache/commons/logging/impl/WeakHashtable.java

- robert

Mime
View raw message