Sanne Grinovero

> The transport should be optional and only used when the jar is present on the classpath and native epoll is not present Haven't checked the latest version of the...

+1 for the approach. I'm also not familiar with the details so I'll stay out of them :)

yeah having a static mutable global state would be quite horrible; the goal here being to cleanup some hacks, not to add more :)

up :)

Awesome - I'm trying to regularly test Quarkus / Hibernate Reactive on a combination of different options, including io_uring, and if the patch could be included in vert.x it would...

We'll need to check why connections aren't released w/o a Transaction - I suspect some scoping issue. But FYI the stated reason to avoid a transaction for performance is misleading,...

Ok that's a good point @OleKsimov - thanks for clarifying your issue. However what you are describing is a starvation of the pool, which is not necessarily related to using...

Thanks for the great reproducer - I finally had some time to dive into it, apologies it took some time. I believe we've indeed found a leak in Hibernate ORM;...

I was wrong yesterday, ORM isn't leaking in practice as there's a second line of defence - need to keep looking.

> We'll need to check why connections aren't released w/o a Transaction - I suspect some scoping issue. It turns out my very initial thought ^ was actually spot on....