Adam Gent

> Some situations/frameworks (e.g. JPA) require a no-args constructor for initialization/set the object after the constructor. I will just add to what @cpovirk wrote is that I suspect (assuming JSpecify...

@xenoterracide You put all your JPA entities in a separate jar/project/module and do not run null analysis on it. Keep the JPA code with packages annotated with `@NullMarked` and continue...

It would be nice but I consider JSpecify at least for its first version more about libraries communicating nullness and less about correct static analysis. What @xenoterracide can largely be...

> I mean I could give you a link to the code if you want to show me how you would make it compile, and run the tests successfully... Without...

> yeah, I mean not validating code is an option. Mine are already/always separated. In this case I think I decided that SuppressWarnings("NullAway") on the getter's is the right solution....

> FYI, "ALL we are doing is adding a nullness axis to the type system" was a foundational decision for this project, though a pointer to that discussion would take...

@xenoterracide > CGLIB is dead, it doesn't work past LTS Java 11 (or was it 17). Java Config can just use proxies... and reflection, which are only magical up to...

yes I am behind on the current state. It appears byte-buddy is being used by hibernate. Regardless whatever the library it uses it still is generating byte code at runtime...

As an exercise I urge you go try to annotate your JPA entities without changing the code with Checker Framework. Then after perhaps propose a solution to solve the JPA...

@netdpb @kevinb9n @cpovirk This is sort of orthogonal to this bug/issue but have you guys made it clear where to file bugs for the reference checker vs the spec itself?...