Serde no_std compatibility #288
Conversation
|
Weird failure in the MSRV test. Wonder if it's a nightly bug in the unused import diagnostic. @semenov-vladyslav maybe try pinning a slightly older nightly? |
|
Looks like clippy failed on unused imports. Forgot to update cfg pragma to a few other imports. Essentially, "serde" feature requires one of "std" or "alloc". Fixed that, now it should work. |
|
Looks like we didn't use We should probably add that: Adding a test here to CI: #289 - checking that this fails as expected. |
|
Confirming no_std build works in main without serde: However with serde it fails as expected: Can we re-base this when the test PR in CI so we ensure in CI it's fixed. |
as an alternative to dalek-cryptography#288 this updates serde `Serialize` and `Deserialize` implementations to use a custom visitor, removing the need for `alloc` or `std` for embedded use, and making this consistent with implementations in [curve25519-dalek](https://github.com/dalek-cryptography/curve25519-dalek/blob/a63e14f4ded078d6bf262ba0b3f47026bdd7f7c0/src/edwards.rs#L269). @pinkforest seems like it'd be good to have some serde tests / this should go over dalek-cryptography#289? Co-Authored-By: Vlad Semenov <[email protected]>
…lize` and `Deserialize` implementations to use a custom visitor, removing the need for `alloc` or `std` for embedded use, and making this consistent with implementations in [curve25519-dalek](https://github.com/dalek-cryptography/curve25519-dalek/blob/a63e14f4ded078d6bf262ba0b3f47026bdd7f7c0/src/edwards.rs#L269). @pinkforest seems like it'd be good to have some serde tests / this should go over dalek-cryptography#289? Co-Authored-By: Vlad Semenov <[email protected]>
This PR addresses issue #287. Tests pass with different combinations of features: "serde", "serde alloc", "serde std", "alloc".