All APIs like env.new_*_array() should return a JObject wrapper (like JByteArray) allowing safe JValue conversion

[EliseChouleur]

Particulary to be able to call a methode with a byte array in parameter.
It was possible in JNI 0.19 thanks to into but the new conversion methode doesn't work with jbyteArray.

e.g:
jni-rs 0.19

    let buffer = env.new_byte_array(asset_size)?; 
    env.call_method(asset_stream, "read", "([B)I", &[buffer.into()])?;

[rib]

Ah, right, I guess this was affected when addressing the soundness issues with allowing safe conversions from sys types: #197

The odd thing here is that jni-rs is directly returning a sys type from env.new_byte_array instead of a type like JObject that has an associated lifetime for the local reference.

We should also track this in the context of #392 where we're looking to ensure that local reference types are non-Copy.

Since the array is effectively just an object you can pass your array to method calls like this for now:

    let buffer = env.new_byte_array(asset_size)?;
    let buffer_obj = unsafe { JObject::from_raw(buffer) };
    env.call_method(stream, "read", "([B)I", &[buffer_obj.into()])?;

We probably shouldn't directly allow safe conversion between jbyteArray and JValue - similar to not allowing safe conversions of jobject to JObject otherwise safe code can can easily pass literal numbers as an invalid pointers.

What we probably need to do here is introduce a JByteArray<'env> type similar to JString or JClass that would be returned by env.new_byte_array() and then there should be a safe way to convert a JByteArray into a JValue (which should make your original code work again)

[EliseChouleur]

Hello
Thanks for the temporary fix !