Initialization Tasks failing

[frootshoot]

Describe the bug

We have been using Liquibase with Kubernetes where in our own helm charts we are configuring a job to run with "QUARKUS_INIT_AND_EXIT" to execute the schema migration.

This was working up to version 3.8.x, and now fails with newer versions due to a NullPointerException on startup - this stops the deployment progressing.

2024-06-11 11:57:55,590 ERROR [io.qua.run.shu.ShutdownRecorder] (Quarkus Main Thread) Graceful shutdown failed: java.lang.NullPointerException: Cannot invoke "java.util.List.size()" because "io.quarkus.runtime.shutdown.ShutdownRecorder.shutdownListeners" is null
	at io.quarkus.runtime.shutdown.ShutdownRecorder.executePreShutdown(ShutdownRecorder.java:42)
	at io.quarkus.runtime.shutdown.ShutdownRecorder.runShutdown(ShutdownRecorder.java:33)
	at io.quarkus.runtime.Application.stop(Application.java:207)
	at io.quarkus.runtime.Application.stop(Application.java:155)
	at io.quarkus.runtime.ApplicationLifecycleManager.run(ApplicationLifecycleManager.java:211)
	at io.quarkus.runtime.Quarkus.run(Quarkus.java:71)

Following an investigation, it appears that the graceful shutdown code was refactored after 3.8.x which removed the null check for shutdown listeners - however it appears any shutdown listeners are not actually injected until after the point when "init and exit" has finished, leading to the NPE.

Expected behavior

The Quarkus container should run the Liquibase migration and then exit cleanly.

Actual behavior

The Liquibase migration happens as expected but the contianer does not shutdown without the error - stopping the appplication containers from starting.

How to Reproduce?

Running any Quarkus application with environment variable "QUARKUS_INIT_AND_EXIT" set to true.

Output of uname -a or ver

No response

Output of java -version

No response

Quarkus version or git rev

3.11.1

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

No response

Additional information

No response

[quarkus-bot]

/cc @geoand (kubernetes), @iocanel (kubernetes)

[geoand]

Thanks for reporting!

As you've already investigated this, would you like to contribute a fix?

[frootshoot]

I am not sure of the intention of the refactor, as it clearly removed the null check - i.e. if the shutdown listeners are supposed to be injected by this point?

[geoand]

What likely happened is that when things were refactored, this use case was not taken into account

[frootshoot]

As you've already investigated this, would you like to contribute a fix?

I am currently not in a position to contribute code, but happy to help where I can with this.

[geoand]

No problem!

#41117 is the fix