One question, has ur usecase ever worked before? Wondering whether this is a regression or just uncovered case for those features

I'd have to check as well: I think the idea would be that root wrapping/unwrapping should not be done for case of converting values, including valueToTree().
But it is definitely bit of a grey area.

@stickfigure Small thing tho, in any case, modifying your reproduction to be decompiled version(plain Java without @Value) would be great! 👍🏼

I'm pretty horrified by WRAP_ROOT_VALUE but someone in this company started with it a decade ago and I'm kinda stuck with it. I have no idea if valueToTree() ever worked in this context, I only inherited this codebase recently.

In the short run I've subclassed ObjectMapper and overridden valueToTree() with something approximately like this, which makes the method work the way I would expect:

return (T)normalNonWrappingObjectMapper.readTree(writeValueAsString(object));

I changed my sample to use record - sorry.

It's hard for me to imagine any situation where someone wouldn't expect valueToTree() to output the equivalent of writeValueAsString(). Seems like that should be an invariant.

It's hard for me to imagine any situation where someone wouldn't expect valueToTree() to output the equivalent of writeValueAsString(). Seems like that should be an invariant.

If mapper.readTree(writeValueAsString) works, that's valid point in terms of intuition from naming itself. But then again, there may not be complete symmetry between plain POJO handling vs ObjectNode(or sorts).
TODO: verify mapper.readTree(writeValueAsString) works.

Lemme trying to read documentation (shared below) see what direction we should be heading.

Yeah, JsonNode is expected to be faithful representation of content exactly as-is.
I may be misremembering aspect of WRAP_ROOT_VALUE not being enabled.. maybe I'll need to dig in to see.
(I vaguely remember something being problematic wrt wrap/unwrap having to be balanced but can't recall details now).

Ok: wrapping is definitely disabled for convertValue():

    protected Object _convert(Object fromValue, JavaType toValueType)
        throws IllegalArgumentException
    {
        // inlined 'writeValue' with minor changes:
        // first: disable wrapping when writing
        final SerializationConfig config = getSerializationConfig().without(SerializationFeature.WRAP_ROOT_VALUE);
        ...

Ah ha. There is some prior art -> #4047.

Funny I didn't even remember this, despite @JooHyukKim and I discussing it just a year ago. :)

Ah. Actually, this might be working as expected -- there is mismatch wrt expected "Root name": on serialization simple class name ("Thing") is used, and expectation is that same is the case when reading (that is, Root Name expectation is set by simple name, JsonNode).

The same thing would occur if explicit ObjectMapper.writeValue() + ObjectMapper.readTree() was used -- although there it'd be possible to use something like:

mapper.reader().withRootName("Thing").readTree(json);

to override expected root name.
In fact, I think that is probably what you need here.

Or... if you want to keep "Thing" as wrapper, simply leave DeserializationFeature.UNWRAP_ROOT_VALUE as false (or set explicitly).

I am not sure there is anything to be done on Jackson side.

Funny I didn't even remember this, despite @JooHyukKim and I discussing it just a year ago. :)

Strange it felt so new to me 😂.

In this case, we may close this as working-as-expected and try see what we can do on documentation side.

Agreed. Solution for @stickfigure is, I think, not enabling UNWRAP_ROOT_VALUE (or disabling it for these reads)