FIX: Recursive batch flush can save beans in wrong order

[rPraml]

if you try to save a persist graph, you normally expect, that the elements are in a logical order.

Here:

• save customer 1
• save contacts a,b,c of customer 1
• save customer 2
• save contacts d,e,f of customer 2
  OR
• save customer 1+2
• save contacts a,b,c of customer 1 + d,e,f of customer 2

This works fine, if no batch flush occur.

Now, it can happen, that during save of customer 1, something (a DB.find in a postInsert hook, a lazy load) triggers a batch flush, which causes to process all internal save queues NOW.
As ebean is currently processing the "customer" query, and the next step of saving customer 2 is interrupted by the flush,
ebean tries to save contact a,b,c + d,e,f - the later three will fail, because customer 2 is not yet saved and has no ID.

As suggested solution, we disable transaction.setFlushOnQuery in the executeNow.

Roland

---

Co-authored-by: Noemi Praml noemi.praml@foconis.de
Co-authored-by: Roland Praml roland.praml@foconis.de