Hi Sanne!

thanks for you quick response.
So, I had not idea, that a bytecode provider can be selected. So we were using defaults.
i.e. in 5.2.17 it was javassist, and in 5.3.7 and 5.4.0-SNAPSHOT it was bytebuddy 1.9.3.

So did some testing and attaching logs with output of mvn compile, the difference is 26sec vs 2mins in 5.2.17 vs 5.3 case.
The 5.4.0-SNAPSHOT is current master ea42f43881fbeced6ba5be413d97e0368e1ccdf and it removes 4 seconds from 5.3.7 time.

So found that property that switches the bytecode provider and when switched 5.4.0-SNAPSHOT to javassit, we are back on 5.2.17 numbers. So indeed it is because of the that provider or some layer in between.

Basically processing with bytebuddy pulls in much more classes (and probably will pull in whole project in the end) and hence the processing is so slow.

Attaching the logs. hibernate_enhance.zip Doing shallow review of the logs. The bytebuddy version when post-processing an Entity prints also post-processing of some enums, interfaces (even not used by that class) and some other code :/
Strange. Do you feel that it is safe to switch to javassist? If so, I need to for both compile time (then it should be used by enhance) and runtime (not sure if it used than).

Any advice would be much appreciated
Thanks!
Tomas