Customize list of exceptions to unwrap in resteasy-reactive

[kevinferrare]

Description

Exceptions handling declared via ServerExceptionMapper annotation are not called when the exception is wrapped into another exception.

Looking at quarkus code, it looks like the RuntimeExceptionMapper class is responsible for unwrapping exceptions, but it only unwraps exceptions that are in a predefined list computed at build time via build items.

Example for resteasy reactive extension:

quarkus/extensions/resteasy-reactive/rest/deployment/src/main/java/io/quarkus/resteasy/reactive/server/deployment/ResteasyReactiveScanningProcessor.java

Line 122 in 6619648

return List.of(new UnwrappedExceptionBuildItem(ArcUndeclaredThrowableException.class),

I propose to add a configuration property to allow users to customize this list of exceptions to unwrap.

Implementation ideas

No response

[quarkus-bot]

/cc @FroMage (resteasy-reactive), @geoand (resteasy-reactive), @stuartwdouglas (resteasy-reactive)

[geoand]

I think this makes sense

[gsmet]

Is it a custom exception that you want to unwrap? Or something that would make sense in general?

[geoand]

It makes sense to have this be configurable.
I planned to do it when someone asked 😀

[gsmet]

Oh yeah, I totally agree but I wanted to know the use case as it might be something we want to add as default anyway for better user experience.

[geoand]

One question I do have however is whether the feature should be enabled by a configuration property or some new annotation.

The reason the configuration property (although totally reasonable) can be suboptimal is that the IDE doesn't provide any assistance or validation of the value

[geoand]

#42094 is what I have in mind

[kevinferrare]

Thank you for the quick and welcoming feedback!
@gsmet

Is it a custom exception that you want to unwrap? Or something that would make sense in general?
It is both, we have java.util.concurrent.CompletionException and a custom exception, thrown by an internal library.

@geoand
After opening the issue, I started to wonder what would be the drawback of also optionally providing the option of just unwrapping all exceptions unconditionally until a handler is found.

In our use case, I don't think this would lead to unwanted behaviours.

[geoand]

After opening the issue, I started to wonder what would be the drawback of also optionally providing the option of just unwrapping all exceptions unconditionally until a handler is found.

Although this might make sense, it is however not how the JAX-RS / Jakarta REST specifies things, so it's best not to make this the default behavior.