@Transactional should not override @RunOnVirtualThread in reasteasy reactive

[imperatorx]

Describe the bug

PR #24942, as choice of caution, disables virtual threads when @Transactional is added to a reasteasy reactive method. Narayana will soon (7.0) have support for virtual threads (jbosstm/narayana#2077), so I think this restriction should go away for those who want to experiment with or build on this.

With narayana built from trunk and a loom friendly jdbc driver, @RunOnVirtualThread works if the method is not annotated with @Transactional, but there is programmatic transaction control inside using QuarkusTransaction.requiringNew().

Expected behavior

@Transactional
@RunOnVirtualThread
public String foo() {
  // access SQL datasource or cusom XA resouce here works, transaction is present
  // is executed on a virtual thread
}

Actual behavior

@Transactional
@RunOnVirtualThread
public String foo() {
  // access SQL datasource or cusom XA resouce here works, transaction is present
  // is executed on the main thread pool instead of a virtual thread
}

How to Reproduce?

No response

Output of uname -a or ver

No response

Output of java -version

19

GraalVM version (if different from Java)

No response

Quarkus version or git rev

3.1.2.Final

Build tool (ie. output of mvnw --version or gradlew --version)

No response

Additional information

No response

[geoand]

cc @cescoffier @franz1981

[franz1981]

AFAIK narayana is now pinnnig the virtual thread; maybe a fix is in flight for that.

[geoand]

That is what I think the issue is saying

[imperatorx]

AFAIK narayana is now pinnnig the virtual thread; maybe a fix is in flight for that.

It is, I made the pr (merged):

Narayana will soon (7.0) have support for virtual threads (jbosstm/narayana#2077),

[cescoffier]

Yes, it got fixed last week. So we can finally improve.

Do you know when we would be able to update narayana? Because that's what will drive the change.

[imperatorx]

Do you know when we would be able to update narayana? Because that's what will drive the change.

They said it will be in 7.x as some public methods got their synchronized flags removed (public API change)

[imperatorx]

Also, what are the plans regarding virtual threads and classic resteasy integration? It seems virtual threads are a better fit for blocking (non-reactive) semantics.

[mmusgrov]

The plan is to release Narayana with @imperatorx 's enhancement this week.

[imperatorx]

7.0.0.Final was released: https://search.maven.org/search?q=g:org.jboss.narayana.jta

[mmusgrov]

The quarkus component upgrade PR has nearly completed its checks.

[cescoffier]

I just merged the narayana update - will lift the limitation tomorrow

[cescoffier]

Fixed with #34808