Inconsistent behaviour with .as_slice()

[kevaundray]

Aim

The double_verify_proof was recently not working as expected for barretenberg.

The code that panics in barretenberg:

use dep::std;

fn main(
    verification_key : [Field; 114], 
    proof : [Field; 94], 
    public_inputs : [Field; 1], 
    key_hash : Field, 
    input_aggregation_object : [Field; 16],
    proof_b : [Field; 94],
) -> pub [Field; 16] {
    let output_aggregation_object_a = std::verify_proof(
        verification_key.as_slice(), 
        proof.as_slice(), 
        public_inputs.as_slice(), 
        key_hash, 
        input_aggregation_object
    );

    let output_aggregation_object = std::verify_proof(
        verification_key.as_slice(),
        proof_b.as_slice(),
        public_inputs.as_slice(),
        key_hash,
        output_aggregation_object_a
    );

    let mut output = [0; 16];
    for i in 0..16 {
        output[i] = output_aggregation_object[i];
    }
    output
}

The code that works in barretenberg:

use dep::std;

fn main(
    verification_key : [Field; 114], 
    proof : [Field; 94], 
    public_inputs : [Field; 1], 
    key_hash : Field, 
    input_aggregation_object : [Field; 16],
    proof_b : [Field; 94],
) -> pub [Field; 16] {

    let verification_key : [Field] = verification_key;
    let proof : [Field] = proof;
    let proof_b : [Field] = proof_b;
    let public_inputs : [Field] = public_inputs;

    let output_aggregation_object_a = std::verify_proof(
        verification_key, 
        proof, 
        public_inputs, 
        key_hash,
        input_aggregation_object
    );

    let output_aggregation_object = std::verify_proof(
        verification_key,
        proof_b,
        public_inputs,
        key_hash,
        output_aggregation_object_a
    );

    let mut output = [0; 16];
    for i in 0..16 {
        output[i] = output_aggregation_object[i];
    }
    output
}

Expected Behavior

Both of these codepaths should work

Bug

We get a panic in barretenberg and barretenberg when deserializing says that input aggregation object has 1 value instead of 16. It is the witness with index 342.

To Reproduce

Installation Method

None

Nargo Version

No response

Additional Context

No response

Would you like to submit a PR for this Issue?

No

Support Needs

No response

[kevaundray]

Note that both of these pass execution, so it may be a bug in serde-reflect

[kevaundray]

Retroactive questions so that we can spot this earlier:

• Why did this not pop up on any other test?
• Why do both of those pass execution, but only fail in barretenberg?