Re: [eclipselink-users] Map-related Exception in M5

[Tom Ware <tom.ware@xxxxxxxxxx>]

Hi rv,

  I'm glad you found a way to work.  It looks to me like the issue you are 
seeing with flush is related to an observation made by James Sutherland in 
GlassFish bug 3937.  It appears we are missing a bit of clean up after flush in 
TopLink Essentials.  It is likely the same issue is occuring in EclipseLink.

  If you'd like you can enter an EclipseLink bug describing your issue.

Hopefully everything is going well now,
Tom

wintermute@xxxxxxxxxxxxxxxxxx wrote:
> Hello Tom,
>
> >   Does adding an em.persist(data) some time before your flush help?
>
> Unfortunately, no. Actually, em.persist(data) was there when I first 
> encountered the exception, before I started to play around, trying to 
> reduce sources of error. With or without the persist(): NPE.
>
> But:
>
> I just removed the flush(), and everything works fine now. The flush() 
> had been there because Essentials (in our development setup) did not 
> update the ID field after a persist() call. I think it should (it's in 
> the spec, isn't it?), and there is a GF bug for this, #3389 to be precise.
>
> But EclipseLink does update the ID, of course. So, no more flush()es, no 
> more pain. I knew flush()es were evil, but I didn't know they were THAT 
> evil.
> I'll go throw out all the flush() calls now and check if that solves the 
> caching problems I was seeing some weeks ago.
>
> Thanks for listening,
>
> rv
> _______________________________________________
> eclipselink-users mailing list
> eclipselink-users@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/eclipselink-users