The CPU pegs at 100% and loading the binary takes about 10-30s. For our tests this adds up. See attached profiling screenshot showing about 2:30 minutes of measure time.
This is a regression and did not happen on version 5.1.0 and 5.2.0 But happens on 5.1.1, 5.1.2, 5.2.1, 5.2.2
One more data point, the CPU pegging is not there in 5.2.0 but present in 5.2.1
, could you give 5.0.11 a try as well, to see if that version has the same issue?
Unfortunately our code now does not work with 5.0 version, some other issue (I forget what I had to change when I migrated to 5.1 and 5.2)
, could you add an attachment similar to the one you attached, except with PersisterFactoryImpl#createEntityPersister and SessionFactoryImpl#checkNamedQueries expanded as many levels as necessary to see which methods are responsible for the increase?
It would also be very helpful to see the same thing for a previous version that does not have the regression.
What would be most helpful would be a test case that reproduces this, or the result of `git bisect` that identifies the commit that causes the change.
I have the entire trace exported as html in the attached archive. Most of the stuff points to building HQL graph and lot of Java class loader and reflection stuff. I suspect a lot of these can be optimized.
I am not sure how to create a repro as this seem to be data dependent and our schema/java code is proprietary. I will have jump through a lot of hoops to get it public, not sure if I will be able to do that within any reasonable turn-around time for debugging.