`StackOverflowError` in Dynamic `StdKeySerializer`

There seem to be a problem (checked and doesn’t seem to be fixed in latest version) with the serialize method of the Dynamic static class of the StdKeySerializer.

        @Override
        public void serialize(Object value, JsonGenerator g, SerializerProvider provider)
                throws IOException
        {
            Class<?> cls = value.getClass();
            PropertySerializerMap m = _dynamicSerializers;
            JsonSerializer<Object> ser = m.serializerFor(cls);
            if (ser == null) {
                ser = _findAndAddDynamic(m, cls, provider);
            }
            ser.serialize(value, g, provider);
        }

The problem comes from the fact that when ser is null, the new ser returned by _findAndAddDynamic is incorrectly filled.

        protected JsonSerializer<Object> _findAndAddDynamic(PropertySerializerMap map,
                Class<?> type, SerializerProvider provider) throws JsonMappingException
        {
            PropertySerializerMap.SerializerAndMapResult result =
                    // null -> for now we won't keep ref or pass BeanProperty; could change
                    map.findAndAddKeySerializer(type, provider, null);
            // did we get a new map of serializers? If so, start using it
            if (map != result.map) {
                _dynamicSerializers = result.map;
            }
            return result.serializer;
        }

So say we are in ser#1, ser#1._dynamicSerializers now has the correct PropertySerializerMap$Single. However, result.serializer._dynamicSerializers has PropertySerializerMap$Empty. Therefore, a new call with that result ser#2 is made which ends up creating an infinite loop.

Possible fix:

• replace _dynamicSerializers = result.map; by result.serializer._dynamicSerializers = result.map

If I’m mistaken please let me know, but It seems obvious when debugging that something’s is not working as intended

[cowtowncoder]

@kpronovost I think I can figure out this case, but it probably does make sense to use a work-around in the meantime. But I think it’s good to fix this as the reproduction looks like something that would occur quite easily.

[kpronovost]

Thanks for the quick reply.

I already patched with realistic data in all test cases, I linked this issue to our Merge Request and will remove unneeded patch once this is resolved in a later version.

TLDR: Thanks for the fix; not a blocker.