Bugfix/113926 fix invalid jsonnode deserialization

[Maximys]

Current PR fix #113926 .
Please pay attention on next lines:

1. https://github.com/Maximys/runtime/blob/cf9d968699c1427227d3e43e32fac187dd0c88d8/src/libraries/System.Text.Json/src/System/Text/Json/Reader/Utf8JsonReader.TryGet.cs#L1482
2. https://github.com/Maximys/runtime/blame/cf9d968699c1427227d3e43e32fac187dd0c88d8/src/libraries/System.Text.Json/src/System/Text/Json/Reader/Utf8JsonReader.TryGet.cs#L1528
3. https://github.com/Maximys/runtime/blame/cf9d968699c1427227d3e43e32fac187dd0c88d8/src/libraries/System.Text.Json/src/System/Text/Json/Reader/Utf8JsonReader.TryGet.cs#L1648

which was extracted from existed classes:

1. runtime/src/libraries/System.Text.Json/src/System/Text/Json/Serialization/Converters/Value/HalfConverter.cs

   Line 43 in c0c5520

   Span<byte> byteBuffer = bufferLength <= JsonConstants.StackallocByteThreshold

2. runtime/src/libraries/System.Text.Json/src/System/Text/Json/Serialization/Converters/Value/Int128Converter.cs

   Line 41 in c0c5520

   Span<byte> buffer = bufferLength <= JsonConstants.StackallocByteThreshold

3. runtime/src/libraries/System.Text.Json/src/System/Text/Json/Serialization/Converters/Value/UInt128Converter.cs

   Line 41 in c0c5520

   Span<byte> buffer = bufferLength <= JsonConstants.StackallocByteThreshold

If I correctly understand, this lines is dead and represend result of copy-paste. If so, I can delete this condition, but may be I does not understand some corner case.

[dotnet-policy-service]

Tagging subscribers to this area: @dotnet/area-system-text-json, @gregsdennis
See info in area-owners.md if you want to be subscribed.

[huoyaoyuan]

Please avoid including large refactor without first discussing with the owners. It can have many unintentional effects.

[huoyaoyuan]

It's not efficient to create dedicated converter class for each generic instantiation.

Instead, create a converter factory for Generic<T>, and invokes the converter of T. Check how NullableConverter, FSharpOptionConverter and JsonCollectionConverter invokes JsonConverter<TElement>.

[Maximys]

I thought the performance of this approach was better

[huoyaoyuan]

While it may improve performance to inline one level of virtual call, it can be easily compensated by other stuff like the ConcurrentDictionary lookup. It's way not maintainable.

[huoyaoyuan]

I also noticed a behavioral discrepancy. If the caller specifies a converter for primitive type, it will be respected by the "container" converters (collections/nullable), but not JsonObject. Even if we decide to keep this behavior, it should be achieved in generic way, by acquiring corresponding JsonPrimitiveConverter<T>.

I'd suggest to hold off and let the area owner make a decision on this.

[Maximys]

You are right, and this code already removed

[huoyaoyuan]

These changes can be reverted once you are invoking the corresponding JsonConverter<Primitive>.

[Maximys]

Removed by force push 9047de2

[huoyaoyuan]

Do not store converters like this. Use a converter factory to create converter for each JsonValuePrimitive<T>. Check NullableConverterFactory.

[Maximys]

Thanks. Fixed by force push 9047de2

[eiriktsarpalis]

Thanks, and apologies for the delay in responding. I would suggest a simpler solution which would involve making JsonValueConverter a generic type with the T : JsonValue. Then you could just insert JsonValueConverter<JsonValuePrimitive> in the list of converters in the resolver.

[dotnet-policy-service]

This pull request has been automatically marked no-recent-activity because it has not had any activity for 14 days. It will be closed if no further activity occurs within 14 more days. Any new comment (by anyone, not necessarily the author) will remove no-recent-activity.

[dotnet-policy-service]

This pull request will now be closed since it had been marked no-recent-activity but received no further activity in the past 14 days. It is still possible to reopen or comment on the pull request, but please note that it will be locked if it remains inactive for another 30 days.