* Fix quite a lot of caching and related statistics use cases
* Support selecting ToOne- and PluralAttributeMappings in LoaderSelectBuilder
* Fix multiple bag role handling in LoaderSelectBuilder
* Fix dynamic batch id entity loader problems
* Simplify lock handling and cache JdbcSelect for SingleIdLoadPlan
* Add support for, but at the same time deprecation-warn about component mappings with different structure but same Java type. See HHH-14660
* Fix enhancement issues and implement support for property laziness
* Implement association key access for both sides of the FK
* Fix subpart access for nullable + TARGET side ToOneAttributeMapping
* Pass through fetch timing to CircularFetch to respect fetch profiles
* Respect fetch profiles for CollectionLoader and SingleIdLoadPlan
* Support group by and order by for entity and embeddable paths/joins
* Don't apply lock clause for LockMode.READ and less
* Change result type for round to argument type
* Fix inference for tuple parameter elements when comparing against embeddables
a
- fix path replacements (test bundles)
- drop inaccurate attempts to access javadoc/sources jar tasks and outputs for jakarta projects - they are not producing them (alternatively, and eventually will have to, we could have them produce those artifacts)
- skip jakarta modules for aggregated javadoc building
* fixed problem in release.gradle that caused "configuration phase" to take much longer than it should
* parallelized the build
* Initial TREAT tests
* fixed bug in `hibernate-core-jakarta.gradle` related to JPA packaged test bundle handling
The runtime classpath contains the dependencies as jars that are not yet built when the plugin needs them. So use the compile classpath plus the compiled classes of the current project.
The runtime classpath contains the dependencies as jars that are not yet built when the plugin needs them. So use the compile classpath plus the compiled classes of the current project.
- temporarily move `publishAlways` to just CI runs until Gradle can clarify how to use that when users are not authenticated; I was originally told that `publishAlways` would simply skip publishing if credentials are not present, but that is clearly not the case
Also introduce SequenceMismatchStrategy#NONE, which allows to fully disable the checks
performed by SequenceMismatchStrategy on initialization; combining these two
changes, users have the option to fully skip loading the details about existing
sequences from the DB.
