Jordan Zimmerman

https://github.com/Randgalt/record-builder/pull/180

Here's an example. Given the attached classes: ```java Jdbi jdbi = Jdbi.create(...) .installPlugin(new SqlObjectPlugin()) .setTransactionHandler(new MyTransactionHandler(LocalTransactionHandler.binding())); jdbi.useExtension(Dao.class, Dao::selectOneWithAnnotation); jdbi.useTransaction(handle -> handle.attach(Dao.class).selectOneWithAnnotation()); ``` It should print `MyTransactionHandler: inTransaction called` twice but...

We've created an alternate delegating handler and will enforce always using it. If you change the enclosed `MyTransactionHandler` to extend `SafeDelegatingTransactionHandler` then the test works correctly (as do all our...

I'm not sure there's a good answer. At minimum I suggest documenting it. Possibly, include something like `SafeDelegatingTransactionHandler` in the library and deprecate `DelegatingTransactionHandler`.

Yes, it appears to. Thank you.

OK - that works. Why then even have the Guava versions? Also, this required two annotations per interface right?

Sure, that would be useful. A contribution would be appreciated. Although, wouldn't a BOM be to match RecordBuilder's dependencies? You need to define a record builder version property regardless right?

Given the existence of `addClassRetainedGenerated` a new boolean will be easier to add.

> when Tasks go into the input_needed state they again becoming client and server blocking requests waiting on user input This isn't true, however. Look at [SEP-1699](https://github.com/modelcontextprotocol/modelcontextprotocol/pull/1783): > At any...