Sunday, July 27, 2008
Tally-Ho Resumes!
After quite a long pause, Tally-Ho is once again moving along. This weekend I managed to upgrade to Wicket 1.3.4 from 1.2.6. This was much less painful than I had anticipated. It was primarily a matter of updating to the new package names and then fixing the Validators and calls to getModel() and setModel(). At the same time I also updated my development environment at home to Eclipse Ganymede and the latest Java 1.6 JDK. And just for show, I updated to OpenJPA 1.1.0.
One bit of weirdness to note is that data caching doesn't seem to work properly in OpenJPA using Java 1.6 runtime class retransformation. I discovered this the hard way by forgetting to include the -javaagent VM argument when running my unit tests. On 1.5, this didn't have as weird a failure mode... OpenJPA seemed to cope with it. With 1.6, it seems to lose track of object identity, such that when queries are run that should be returning the same object (or at least the newly queried version of an updated object), sometimes different objects or old versions of objects are returned.
The moral of the story is to specify -javaagent with OpenJPA at runtime if you're not doing static instrumentation of your classes. Class retransformation doesn't work properly yet. If I can find some time, I'll create a simple testcase for the OpenJPA folks demonstrating the problem.
Left to go on Tally-Ho is completing the arbitrary content with attachments feature, and then looking at Spring to see if / what can be leveraged to make life a little bit easier/simpler.
In other news, this is my 200th blog post. Woohoo?
One bit of weirdness to note is that data caching doesn't seem to work properly in OpenJPA using Java 1.6 runtime class retransformation. I discovered this the hard way by forgetting to include the -javaagent VM argument when running my unit tests. On 1.5, this didn't have as weird a failure mode... OpenJPA seemed to cope with it. With 1.6, it seems to lose track of object identity, such that when queries are run that should be returning the same object (or at least the newly queried version of an updated object), sometimes different objects or old versions of objects are returned.
The moral of the story is to specify -javaagent with OpenJPA at runtime if you're not doing static instrumentation of your classes. Class retransformation doesn't work properly yet. If I can find some time, I'll create a simple testcase for the OpenJPA folks demonstrating the problem.
Left to go on Tally-Ho is completing the arbitrary content with attachments feature, and then looking at Spring to see if / what can be leveraged to make life a little bit easier/simpler.
In other news, this is my 200th blog post. Woohoo?
Labels: tally-ho
Subscribe to Posts [Atom]
