Skip to main content
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [eclipselink-dev] Fix for Bug 312947 ready for review
Hi Andrei,

    Unfortunately you haven't seen the Gemini JPA code that relies on these EclipseLink changes because Mike Keith is waiting for these changes to show up in EclipseLink before we apply the Gemini updates.

On 12/7/2010 4:15 PM, Andrei Ilitchev wrote:
JPAInitializer: why to add an extra parameter to registerTransformer (Map properties)? The only use of the existing parameter puInfo is to print out puName - for FINER logging.
In Gemini OSGi JPA I need to pass properties down to the transformer, specifically the Bundle name and version.
EntityManagerFactoryProvider: both initialEmSetupImpls and initialPuInfos are used only in JavaSE case (not OSGi) - why both their class and initial values are changed? They should not be.
They are referenced in PersistenceProvider which is not just SE.  The Gemini code path will hit them.
HashMap is used so that get doesn't lock the map - Hashtable's locking get method would cause performance deterioration.
So we use a highly concurrent map, but truth be told this code is not highly contended.
There is no concurrency concerns for initialPuInfos - it's used (if used at all) only once during initialization;
Again, referenced in PersistenceProvider and in Gemini initialization happens as bundles come and go.
Possible concurrency issues for initialEmSetupImpls resolved by synchronized (EntityManagerFactoryProvider.emSetupImpls) statement in PersistenceProvider (one for put, another for remove).
I think ConcurrentHashMap is a more concurrent solution.

Initial values should stay null for both maps, too:
neither of the maps is used outside of JavaSE case - no reason to always initialize them.
Again, PersistenceProvider and the Gemini code is not JavaSE, it's multi-threaded OSGi.  It's easier to initialize them statically which is thread safe than to write multi-thread aware initialization code.
Even in JavaSE case initialPuInfos only used if the javaagent is used - initializing it application managed case on app. server would mean that changes made to redeployed persistence unit would be ignored.
Gemini provides an OSGi "Agent" that may run multiple times for different bundles to perform initialization and support weaving. 

But could you clarify the issue with redeployment?

    Shaun


On 12/07/2010 3:15 PM, Shaun Smith wrote:
Bug 312947 - Allow EclipseLink JPA to work with Gemini in OSGi
https://bugs.eclipse.org/bugs/show_bug.cgi?id=312947

Minor adjustments to JPA initialization to work with Gemini JPA.  This revised patch does not remove what seemed to be dead temploader code as James advised is still necessary to support weaving legacy apps using sessions.xml.
--
Oracle
Shaun Smith | Principal Product Manager
Phone: +19055023094
Oracle Server Technologies, Oracle TopLink
ORACLE Canada | 110 Matheson Boulevard West, Suite 100, Mississauga, Ontario | L5R 3P4
Green Oracle Oracle is committed to developing practices and products that help protect the environment 

_______________________________________________
eclipselink-dev mailing list
eclipselink-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipselink-dev
  

_______________________________________________
eclipselink-dev mailing list
eclipselink-dev@xxxxxxxxxxx
https://dev.eclipse.org/mailman/listinfo/eclipselink-dev

--
Oracle
Shaun Smith | Principal Product Manager
Phone: +19055023094
Oracle Server Technologies, Oracle TopLink
ORACLE Canada | 110 Matheson Boulevard West, Suite 100, Mississauga, Ontario | L5R 3P4
Green Oracle Oracle is committed to developing practices and products that help protect the environment

Back to the top