key: Improve unit tests #602
Conversation
Improve the unit tests in the `key` module by doing: - Remove a bunch of "rand-std" (use hard coded keys instead) - Split big tests into smaller appropriately named tests - Move things around so we can better see whats being tested
|
I got this idea for a post you made someplace or other about not having tests when "rand-std" feature was disabled. If this PR is good we can probably mimic it for a bunch of other modules. |
| 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, | ||
| 0xff, 0xff, 0x00, 0x00, 0xff, 0xff, 0x00, 0x00, | ||
| 0x63, 0x63, 0x63, 0x63, 0x63, 0x63, 0x63, 0x63, | ||
| ]; |
There was a problem hiding this comment.
Could also use hex_lit::hex!
|
|
||
| let mut pk_bytes = [0u8; 33]; | ||
| pk_bytes[0] = 0x02; // Use positive Y co-ordinate. | ||
| pk_bytes[1..].clone_from_slice(&PK_BYTES); |
There was a problem hiding this comment.
I'm surprised clippy doesn't scream about copy.
| PublicKey::from_slice(&[0; constants::UNCOMPRESSED_PUBLIC_KEY_SIZE + 1]), | ||
| Err(InvalidPublicKey) | ||
| SecretKey::from_slice(&[0; constants::SECRET_KEY_SIZE + 1]), | ||
| Err(InvalidSecretKey) |
There was a problem hiding this comment.
This test isn't that great since all-zero keys are also invalid.
There was a problem hiding this comment.
Nice observation, I'll improve it.
| assert!(kp.eq_fast_unstable(&kp2)); | ||
| } | ||
|
|
||
| // FIXME(Tobin): I don't know what this is testing? There is no assertion or panic? |
There was a problem hiding this comment.
I believe this tests that if an RNG generates an out-of-range key it is called again instead of panicking.
| let (sk, pk) = s.generate_keypair(&mut rand::thread_rng()); | ||
| assert_eq!(PublicKey::from_secret_key(&s, &sk), pk); // Sanity check. | ||
|
|
||
| // TODO: This would be better tested with a _lot_ of different tweaks. |
There was a problem hiding this comment.
According to this comment, maybe we should keep it random?
There was a problem hiding this comment.
Implying that running the tests multiple times improves coverage? The test space (a random Scalar) is so large that no amount of test re-runs in any way improves coverage so I think the use of random gives a false sense of coverage. I've not worked out how to test this (and the other tests still "rand-std") though, I tried playing with kani yesterday without success. Open to suggestions but I probably just need to play with it some more.
There was a problem hiding this comment.
Good point. Maybe just test a few obvious things that are generally edge-cases - all zeroes, max...
|
Maybe we should instead have a |
That's a cool idea. |
|
Thanks for the reviews, I didn't respond to all your comments @Kixunil but I read them and I'll come back to this PR when I'm in a "psyched for testing" mood. |
|
Seems "psyched for testing" is taking a while to materialise, converting to draft. |
Improve the unit tests in the
keymodule by doing: