NullPointerException in McpClientSession

[vorburger]

Please do a quick search on GitHub issues first, there might be already a duplicate issue for the one you are about to create.
If the bug is trivial, just go ahead and create the issue. Otherwise, please take a few moments and fill in the following sections:

Bug description

In enola-dev/enola#1862, I've run into:

java.lang.NullPointerException: data
	at java.base/java.util.Objects.requireNonNull(Objects.java:259)
	at reactor.core.publisher.FluxHandle$HandleSubscriber.next(FluxHandle.java:251)
	at io.modelcontextprotocol.spec.McpClientSession.lambda$sendRequest$12(McpClientSession.java:280)
	at reactor.core.publisher.FluxHandle$HandleSubscriber.onNext(FluxHandle.java:113)
	at reactor.core.publisher.SerializedSubscriber.onNext(SerializedSubscriber.java:99)
	at reactor.core.publisher.FluxTimeout$TimeoutMainSubscriber.onNext(FluxTimeout.java:181)
	at reactor.core.publisher.MonoCreate$DefaultMonoSink.success(MonoCreate.java:176)

I guess that this means that this is null?

Environment

0.14.1

Steps to reproduce

Unfortunately I'm unable to reproduce it locally even myself, but I am consistently seeing this on a failing CI build.

Expected behavior

Clear error message instead of NPE.

Minimal Complete Reproducible example

N/A

[vorburger]

FTR: I'm still seeing this NPE after upgrading from 0.14.1 to 0.15.0. I'll try to debug it using #607. @tzolov

[vorburger]

Running my enola-dev/enola#1862 with #607 I get:

java.lang.IllegalStateException: Failed to unmarshal response: JSONRPCResponse[jsonrpc=2.0, id=4937ec33-1, result=null, error=null]
        at io.modelcontextprotocol.spec.McpClientSession.lambda$sendRequest$12(McpClientSession.java:293)

So it look like there appear to be some cases when that jsonRpcResponse is not itself null, but has neither a result nor an error, which I guess is surprising? Perhaps that should be caught earlier?

[vorburger]

So it look like there appear to be some cases when that jsonRpcResponse is not itself null, but has neither a result nor an error, which I guess is surprising? Perhaps that should be caught earlier?

In my case, this happened because I used my own custom Jackson ObjectMapper.

enola-dev/enola#1904 works around this in my project ("downstream"). But it would probably be good if this library wouldn't just throw a confusing internal NPE, but would catch this problem more clearly and earlier. I'll see if I contribute to that.

[vorburger]

It would probably be good if this library wouldn't just throw a confusing internal NPE, but would catch this problem more clearly and earlier. I'll see if I contribute to that.

#658 proposed for this.