Hibernate ORM - Assorted improvements to proxy generation

[gsmet]

Better reviewed commit per commit.

[Sanne]

Very nice.

I noticed something odd though - the parameter proxyInterfaceNames passed to buildUnloadedProxy is a Set whose content is always the same? IFF that's the case then the buildUnloadedProxy could resolve the types in advance and also reuse as a constant.
But it looks like suspicious, like if we lost something during previous changes.

[gsmet]

Maybe. I don't want to open a can of worms here.

I tried to "optimize" things a bit here but it makes the code less readable. Given the description of the type will be in the TypePool cache after the first call, I think we are good.

[yrodiere]

The quirkiness is a result of my (botched?) upgrade to Hibernate ORM 7: 6f4cfac#diff-583c184a7bb985fc15d72cd11861e77f12dbd3332307065f3122b8d6390c1602L1371-R1367

In short, previously you could add custom interface names with @Proxy, but that's gone in ORM 7+.

I'll send a PR to do what Sanne suggested.

[Sanne]

nice, thanks 👍

[yrodiere]

Most changes look fine, but unless I'm missing something (and I very well might be), it seems the proxy cache is no longer populated?

This raises the question: was this done to fix a particular problem, and is there a benchmark we could use to check how effective it is? It's fine if there isn't, but if there is that would give a lot more confidence there isn't some hard-to-catch issue like this.

[quarkus-bot]

Status for workflow Quarkus CI

This is the status report for running Quarkus CI on commit 8bd0f55.

✅ The latest workflow run for the pull request has completed successfully.

It should be safe to merge provided you have a look at the other checks in the summary.

You can consult the Develocity build scans.