NullPointerException when compiled to Spring Native binary

[rj93]

Describe the bug
I am using Spring Shell to create a cli tool, the library works fine when run from the jar, however when compiled to a native binary and ran it throws a NullPointerException.

I have added the client-java-spring-aot-integrations dependency but it doesn't seem to have resolved anything, and I am also required to add the following to reflect-config.json otherwise I get a different error.

[
  {
    "name": "io.kubernetes.client.openapi.models.",
    "unsafeAllocated": true
  }
]

When running as native it errors after Kubectl.get(...) is invoked:

2024-04-02T23:40:30.399+01:00  INFO 96712 --- [           main] com.rj93.cli.CliApplication              : Started CliApplication in 0.071 seconds (process running for 0.097)
2024-04-02T23:40:30.400+01:00  INFO 96712 --- [           main] com.rj93.cli.commands.GetPodsCommand     : Getting pods
2024-04-02T23:36:50.127+01:00 ERROR 96399 --- [           main] io.kubernetes.client.util.ModelMapper    : Unexpected exception while loading classes

java.lang.NullPointerException: null
        at io.kubernetes.client.util.ModelMapper.getClassNamesFromPackage(ModelMapper.java:479) ~[na:na]
        at io.kubernetes.client.util.ModelMapper.initModelMap(ModelMapper.java:407) ~[na:na]
        at io.kubernetes.client.util.ModelMapper.<clinit>(ModelMapper.java:81) ~[na:na]
        at io.kubernetes.client.extended.kubectl.Kubectl$ApiClientBuilder.refreshDiscovery(Kubectl.java:236) ~[k8s-cli-native-example:na]
        at io.kubernetes.client.extended.kubectl.KubectlGet.execute(KubectlGet.java:60) ~[k8s-cli-native-example:na]
        at com.rj93.cli.commands.GetPodsCommand.getConfigMaps(GetPodsCommand.java:27) ~[k8s-cli-native-example:na]

Client Version
20.0.1

Kubernetes Version
1.29.1

Java Version
21.0.2-tem

GraalVM Version
21.0.2-graalce

To Reproduce
Reproducable example can be found here: https://github.com/rj93/k8s-cli-native-example

Expected behavior
A NullPointerExpcetion not to be thrown

KubeConfig

apiVersion: v1
clusters:
- cluster:
    certificate-authority-data: <<REDACTED>>
    server: https://kubernetes.docker.internal:6443
  name: docker-desktop
contexts:
- context:
    cluster: docker-desktop
    user: docker-desktop
  name: docker-desktop
current-context: docker-desktop
kind: Config
preferences: {}
users:
- name: docker-desktop
  user:
    client-certificate-data: <<REDACTED>>
    client-key-data: <<REDACTED>>

Server:

• OS: MacOS
• Environment: bash
• K8s Cluster: Docker Desktop

[rj93]

@joshlong I see you implemented the spring-aot module in #2457, would you maybe have an idea what is happening here?

[rj93]

Done some digging and (sort of) found the issue, registering the Spring hints relies on the API model classes being annnotated with @ApiModel, however it appears that they no longer have the annotation so it is returning 0 models to register.

Updating this to find all classes in the io.kubernetes.client.openapi.models package correctly registers all models for reflection:

Reflections reflections = new Reflections("io.kubernetes.client.openapi.models", Scanners.SubTypes.filterResultsBy(s -> true));
Set<Class<?>> classes = reflections.getSubTypesOf(Object.class);
LOGGER.info("Found {} apiModels", classes.size());
for (Class<?> clazz : classes) {
     LOGGER.info("registering {} for reflection", clazz);
    hints.reflection().registerType(clazz, allMemberCategories);
}

2024-04-03T13:58:31.395+01:00  INFO 17552 --- [           main] org.reflections.Reflections              : Reflections took 272 ms to scan 2 urls, producing 565 keys and 5623 values
2024-04-03T13:58:31.489+01:00  INFO 17552 --- [           main] tesBeanFactoryInitializationAotProcessor : Found 1115 models
2024-04-03T13:58:31.489+01:00  INFO 17552 --- [           main] tesBeanFactoryInitializationAotProcessor : registering class io.kubernetes.client.openapi.models.V1beta1Variable$CustomTypeAdapterFactory for reflection
2024-04-03T13:58:31.489+01:00  INFO 17552 --- [           main] tesBeanFactoryInitializationAotProcessor : registering class io.kubernetes.client.openapi.models.V1Endpoint for reflection
...

This appears to fix the use of the *Api classes from the io.kubernetes.client.openapi.apis package, however Kubectl still fails with the original error.

[cache-sk]

This is also mentioned in issue #3134

[k8s-triage-robot]

The Kubernetes project currently lacks enough contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue as fresh with /remove-lifecycle stale
• Close this issue with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale

[cache-sk]

/remove-lifecycle stale

[k8s-triage-robot]

The Kubernetes project currently lacks enough contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue as fresh with /remove-lifecycle stale
• Close this issue with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale

[cache-sk]

Thi should be soon solved thanks to kubernetes-client/gen#270

/remove-lifecycle stale