Introduce Arbitrary crate and add PublicKey arbitrary impl #826
Conversation
shinghim
force-pushed
the
arbitrary
branch
3 times, most recently
from
e025ffe to
e68d976
Compare
src/key.rs
Outdated
| #[cfg(feature = "global-context")] | ||
| impl<'a> Arbitrary<'a> for PublicKey { | ||
| fn arbitrary(u: &mut Unstructured<'a>) -> arbitrary::Result<Self> { | ||
| Ok(PublicKey::from_secret_key(&SECP256K1, &SecretKey::arbitrary(u)?)) |
There was a problem hiding this comment.
In e68d976:
We can be much faster (at least 100x) than this by choosing 32 random bytes, putting 2 or 3 in front of it, and then attempting to deserialize the resulting 33-byte array as a public key. If we fail, we should retry. On average this will take 2 tries.
If the intention is to use this logic in fuzzers and such then we should definitely do this.
There was a problem hiding this comment.
But concept ACK bringing arbitrary into this crate.
There was a problem hiding this comment.
We can be much faster (at least 100x) than this by choosing 32 random bytes, putting 2 or 3 in front of it, and then attempting to deserialize the resulting 33-byte array as a public key. If we fail, we should retry. On average this will take 2 tries.
Nice! Updated with that logic
|
Nice. fa8af6e looks great. I'm a liiiitle tempted to try to save on entropy by using a PRNG for retries. But I think that's the wrong approach. Aside from being slow or biased (or both :)) it's not very kind to fuzzers to try to "stretch" their entropy with a PRNG. Since then the fuzzer wastes effort trying to probe the insides of the PRNG instead of the actual code being tested. |
I was looking into writing some
Arbitraryimpls in thebitcoincrate, and one of the types contained aPublicKey. Any interest in adding the arbitrary feature here?