Issue with decoding response in PocketIC integration tests

Developers
1

Hi,

I’m finally trying PocketIC for the first time and I’m encountering an issue in my integration tests: Candid stuff :slight_smile:

I want to test the following function:

fn schnorr_public_key(arg: SchnorrPublicKey) -> SchnorrPublicKeyReply {}

and do the following in my test:

let reply: Result<SignWithSchnorrReply, String> = update(&pic, my_principal,  canister_id, "sign_with_schnorr", encode_one(&payload).unwrap());

pub fn update<T: CandidType + for<'de> Deserialize<'de>>(
    ic: &PocketIc,
    sender: Principal,
    receiver: Principal,
    method: &str,
    args: Vec<u8>,
) -> Result<T, String> {
    match ic.update_call(receiver, sender, method, args) {
        Ok(WasmResult::Reply(data)) => decode_one(&data).unwrap(),
        Ok(WasmResult::Reject(error_message)) => Err(error_message.to_string()),
        Err(user_error) => Err(user_error.to_string()),
    }
}

However, the tests fail at decoding the reply:

Feb 06 09:12:28.662 WARN s:/n:/ic_state_manager/ic_state_manager No state available with certification.
thread 'test_sign_with_schnorr' panicked at tests/tests.rs:60:58:
called `Result::unwrap()` on an `Err` value: Custom(input: 4449444c026c01f8c5aeab01016d7b0100_40d0e65fdb6cb0baf7c3cd285d268910233a3c175f2ca9ffbd812fed5bd6d9411cca2541b337bc2d33787059e2958515af11cf600369bb5919d4b011282e9444ba
table: type table0 = record { 359_375_608 : table1 }
type table1 = vec nat8
wire_type: record { 359_375_608 : table1 }, expect_type: variant { Ok : record { signature : vec nat8 }; Err : text }

Caused by:
    0: Fail to decode argument 0 from table0 to variant { Ok : record { signature : vec nat8 }; Err : text }
    1: Subtyping error: Type mismatch at ...src/index.crates.io-6f17d22bba15001f/candid-0.9.11/src/de.rs:877)
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
test test_sign_with_schnorr ... FAILED

Complete code is here: GitHub - domwoe/schnorr_canister: Canister to create schnorr signatures on the Internet Computer.

2

Hi @domwoe, it looks like the bug does not come from PocketIC but rather from supplying a different candid format than what the function expects. In the error message, it looks like the canister expects a variant, but a record was supplied.

3

(signature hashes to 359_375_608, so they’re equivalent)

Basically the thing you’re trying to decode Ok(thing) but you receive thing.

Looking at your code you have

pub struct SignWithSchnorrReply {
    pub signature: Vec<u8>,
}

#[ic_cdk::update]
fn sign_with_schnorr(arg: SignWithSchnorr) -> SignWithSchnorrReply {

So your function returns only thing while in your test you try to decode to Ok(thing)

4

Ah, thanks @Vivienne !

5

(signature hashes to 359_375_608 , so they’re equivalent)

How does one arrive at this value/what hash function are you using here? :slight_smile:

6

I’m using didc hash "signature". No clue what algorithm is behind that :slightly_smiling_face: You can download didc in the candid repo

7

Thanks! Somewhat unrelated, but I’ve always wondered why candid does not just tell you in plaintext what it wants. Why does it show these signatures only?

8

A string costs a lot more bytes than a hash. And if you have the interface definition it is easy to reverse