662843e73b Improved error handling in `Parity` serde impl (Martin Habovstiak)
Pull request description:
* Fixes error message to be according to the trait documentation
* Uses `unexpected_value` to provide more information about the error
ACKs for top commit:
apoelstra:
ACK 662843e73b
Tree-SHA512: 2506f06305b01793f64818640931d00564334d96a1e0ef00574faacf1ec8733da13fbf91e57e49fa7c9c06587863fe66145f25afae8d8cabe546dd0ecc48caea
Serializing the value as `u8` is more compact but this is a breaking
change.
`Visitor` was renamed to avoid hungarian notation and maybe allow other
integers in the future.
705c9cfbc1 Clarified conversions between `Parity` and integers (Martin Habovstiak)
Pull request description:
This was discussed in https://github.com/rust-bitcoin/rust-secp256k1/pull/390#issuecomment-1033018430
ACKs for top commit:
apoelstra:
ACK 705c9cfbc1
Tree-SHA512: 3ba2ec566099c3c6d1c6f830e4959312b818b8766d924e3d995e6b23bd196ab747cc03d46f494ef451569188b0163f53e3236cacd20bfae9118ee76bcdbc9c02
4c43d5e20f Add custom Debug impl for RecoverableSignature (Tobin Harding)
Pull request description:
Currently when debug printing the `RecoverableSignature` we do so byte by byte, this means that the output differs depending on the endianess of the machine. If instead we serialize the signature in compact form then the output is the same irrespective of the endianess.
With this applied the following two commands now pass:
```
cargo test test_debug_output --features=recovery
```
```
cross test --target powerpc-unknown-linux-gnu test_debug_output --features=recovery
```
Fixes: #375
ACKs for top commit:
apoelstra:
ACK 4c43d5e20f
Tree-SHA512: 073c2e0e23ce41a2b35f1b1193b07a755b726bf565d61e6bcb23b6bdaab31ba3591f31aa92230b07f7dfc018de0401eba09a6858dc261e66dacb331355f40d76
3c9dd2fb32 Fix example dependency list (Tobin Harding)
Pull request description:
Example relies on `rand-std` not plain `rand` dependency.
I do not understand why the following command passes without this patch
applied
```
cargo test --no-default-features --features=std,rand,bitcoin_hashes
```
But if we put the same code in a standalone binary it fails as expected?
Since the running of this test is _unusual_ and it is primarily meant as
an entry point example to the library, remove the mention of "alloc"
feature and just depend upon "std".
Fixes: #395
ACKs for top commit:
apoelstra:
ACK 3c9dd2fb32
Tree-SHA512: 8e7ec7ac846e2916c29b74c7485650e5242ae1141c12c69b50d74efdfee71c11a52cd454231d2a7cdd6f8f683d3ba4369f9bf898a6b9351dc92c2a4e2bd626cd
eb453b8227 Add global context API (Tobin Harding)
3ecb5e41b3 Refactor from_secret_key definition (Tobin Harding)
e2d47a29e2 Remove unnecessary import statement (Tobin Harding)
d79989bc95 Remove erroneous duplicate feature (Tobin Harding)
Pull request description:
Our API often involves a `Secp256k1` parameter, when users enable the `global-context` feature they must then pass `SECP256K1` into these functions. This is kind of clunky since the global is by definition available everywhere.
Make the API more ergonomic for `global-context` builds by adding various API functions/methods that use the global context implicitly.
The first 3 patches are clean up.
Resolves: #330
ACKs for top commit:
apoelstra:
ACK eb453b8227
Tree-SHA512: 21d89a6688c24a7920d48ea92d923889bec2bbe9dc5ed5e33639405be45a50f50022a28dc1f235b8bea850ac39013c7dd24b5aed086ed40f5b259dd44c06433d
Our API often involves a `Secp256k1` parameter, when users enable the
`global-context` feature they must then pass `SECP256K1` into these
functions. This is kind of clunky since the global is by definition
available everywhere.
Make the API more ergonomic for `global-context` builds by adding
various API functions/methods that use the global context implicitly.
The global context is already in scope in tests since we use a glob
import. No clue why Clippy does not warn for this.
Remove unnecessary import statement in test function.
Recently we introduced uniform styling for module docs over in
`rust-bitcoin` repo. We can do the same here but its a bit controversial
because it removes the heading from module docs and every single public
module in rust-secp256k1 uses a heading. Instead we use a full
sentences. Also makes uniform the trailing `//!`.
As is typical in the Rust ecosystem use the third person tense when
documenting functions. E.g.,
```
/// Creates a new Foo.
```
As opposed to
```
/// Create a new Foo.
```
Example relies on `rand-std` not plain `rand` dependency.
I do not understand why the following command passes without this patch
applied
```
cargo test --no-default-features --features=std,rand,bitcoin_hashes
```
But if we put the same code in a standalone binary it fails as expected?
Since the running of this test is _unusual_ and it is primarily meant as
an entry point example to the library, remove the mention of "alloc"
feature and just depend upon "std".
Currently when debug printing the `RecoverableSignature` we do so byte
by byte, this means that the output differs depending on the endianess
of the machine. If instead we serialize the signature in compact form
then the output is the same irrespective of the endianess.
With this applied the following two commands now pass:
```
cargo test test_debug_output --features=recovery
```
cross test --target powerpc-unknown-linux-gnu test_debug_output --features=recovery
```
Fixes: #375
b3503ba148 Add example to SharedSecret (Tobin Harding)
Pull request description:
Currently the rustdoc on `SharedSecret` is wildly incorrect (possibly a cut'n'pasta error).
Fix the rustdoc for `SharedSecret` and add an examples section to assist testing the public API.
Fixes: #249
ACKs for top commit:
apoelstra:
ACK b3503ba148
Tree-SHA512: 650092388099bb415c11ea335ca6b64c90094f1a51ceecc403911316ee62da0279488af6fa66e00ee5269c129f06d4641085f8ab9be91c98d24a7a4449d235c2
8339ca5706 Add documentation guiding users towards randomization (Tobin Harding)
cf1496b64e Add documentation about rand-std feature (Tobin Harding)
1693d51ce7 Randomize context on creation (Tobin Harding)
a0465ea279 Remove feature global-context-less-secure (Tobin Harding)
Pull request description:
Currently it is easy for users to mis-use our API because they may not know that `randomize()` should be called after context creation for maximum defence against side channel attacks.
This PR entails the first two parts of the plan outlined in #388. The commit messages are a bit light of information as to _why_ we are doing this so please see #388 for more context.
In light of @real-or-random's [comment](https://github.com/rust-bitcoin/rust-secp256k1/issues/388#issuecomment-1026613592) about verification contexts the randomization is done in `gen_new` i.e., for _all_ contexts not just signing ones.
Also, I think we should add some docs about exactly _what_ randomization buys the user and what it costs. I do not know exactly what this is, can someone please write a sentence or two that we can include in the docs to `gen_new`?
@TheBlueMatt please review patch 4.
Resolves: #225
**Note**: This is a total re-write of the original PR, most of the discussion below is stale. Of note, the additional API that takes a seed during construction is not implemented here.
ACKs for top commit:
apoelstra:
ACK 8339ca5706
Tree-SHA512: e74fe9a6eaf8ac40e4e06997602006eb8ca95216b5bc6dca3f5f96b5b4d3bf8610d851d8f1ef5c199ab7fbe85b34d162f2ee0073647f45105a486d20d8c0722a
Currently the rustdoc on `SharedSecret` is wildly incorrect (possibly a
cut'n'pasta error).
Fix the rustdoc for `SharedSecret` and add an examples section to assist
testing the public API.
Fixes: 249
Now that we opportunistically randomize the context on creation if
`rand-std` is enabled it would be nice to encourage users who do not
wish to use `rand-std` to randomize the context. We already have an API
to do this but it requires a separate call to do so. Instead of adding a
bunch of additional constructors elect to add documentation to the
current constructors guiding users towards randomization.
We recently implemented opportunistic randomization of the context
object if the the `rand-std` feature is enabled. Both for the global
context and also for signing context constructors.
Add documentation about `rand-std` feature in relation to the context
object.
Instead of providing a mechanism for users to opt out of randomization
we can just feature gate the call site i.e., opportunistically randomize
the global context on creation if `rand-std` feature is enabled.
2732891359 Change rand to rand-std in lib.rs documentation (Vincent Liao)
Pull request description:
I copy-pasted the key-generation example written on the documentation, but it didn't work. It only worked when I used the feature `rand-std` instead of `rand`.
To reproduce, boot up a new Rust project, and add this to main.rs:
```
use secp256k1::rand::rngs::OsRng;
use secp256k1::{Secp256k1, Message};
use secp256k1::hashes::sha256;
let secp = Secp256k1::new();
let mut rng = OsRng::new().expect("OsRng");
let (secret_key, public_key) = secp.generate_keypair(&mut rng);
let message = Message::from_hashed_data::<sha256::Hash>("Hello World!".as_bytes());
let sig = secp.sign_ecdsa(&message, &secret_key);
assert!(secp.verify_ecdsa(&message, &sig, &public_key).is_ok());
```
Using this dependencies causes error: `secp256k1 = {version="0.21.2", features=["rand", "bitcoin_hashes"]}`. After replacing `rand` with `rand-std`, it works.
ACKs for top commit:
apoelstra:
ACK 2732891
tcharding:
tACK 2732891359
Tree-SHA512: 6b5436bc71bab7535e432e119679bc6bcb11d2575b609e039cc25c122ae92b528f95a673e9c643a6cfa2ee3a663f7efdd61731b6084261c52a220448b6f72d12
c30026d08b Fix typo 'epected' -> 'expected' (Tobin Harding)
f3688ecf56 Use rand-std in key rustdoc examples (Tobin Harding)
ae3e06f95b Fix lint warnings in test code (Tobin Harding)
c01cd8f1f3 Enable running tests without default features (Tobin Harding)
a79840eca2 Be explicit about example feature requirements (Tobin Harding)
433c350424 Add multiple implementations of Debug for secrets (Tobin Harding)
632ecc4530 Use fully qualified path for mem (Tobin Harding)
Pull request description:
As indicated by the comment in `contrib/test.sh` we should be able to test with `--no-default-features`.
- Patch 1 uses fully qualified path to remove a build warning.
- Patch 2 adds additional `Debug` implementations for secrets, uses `bitcoin_hashes` if available, please review carefully.
- Patch 3 adds `std` as an explicit requirement for the three examples
- Patch 4 enables `cargo test --no-default-features, fixes all the feature gating in unit tests.
- Patch 5 fixes lint warnings generated while running the feature matrix in `contrib/test.sh`.
**Please Note**: Currently the `alloc` feature cannot be built with Rust 1.29, this made it into master because we don't build ever with the `alloc` feature enabled in CI. This PR _should_ add `alloc` to the features matrix but it does not. Adds a TODO comment to `contrib/test.sh` to add it once we bump MSRV.
ACKs for top commit:
apoelstra:
ACK c30026d08b
Tree-SHA512: 3bbdda332ab1e04eaa3479d9e9c7463a54347f56019ce5366bb36eb8d5ccaced32539e2c58454a7714d76b7bab9f1ab56accb04de67c826165dd104ac0b3b893
Seems there is a bug in cargo, the tests in `key.rs` run successfully
but AFAICT they should fail. Here is an example, running `cargo test
--features=rand` should make this test fail but it doesn't?
```
/// Secret 256-bit key used as `x` in an ECDSA signature.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// # #[cfg(all(feature = "rand", any(feature = "alloc", feature = "std")))] {
/// use secp256k1::{rand, Secp256k1, SecretKey};
///
/// let secp = Secp256k1::new();
/// let secret_key = SecretKey::new(&mut rand::thread_rng());
/// # }
/// ```
Anywho, use the correct feature gate: `rand-std`.
Various combinations of features trigger lint warnings for unused code,
all warnings are caused by incorrect feature gating.
Correct feature gating to remove Clippy warnings during testing.
Currently various features fail to build when enabled without default
features. This is because many tests need feature gating.
Feature gating the import statements quickly turns into spaghetti when
trying to cover all combinations of two features correctly, instead just
allow unused imports on `tests` modules where needed.
Add correct feature requirements to the examples so they also can be run
without default features.
Improve the CI script by doing:
- Add `std` to the feature matrix.
- Add `--no-default-features` to test runs in the CI script.
The examples depend on having the "std" feature [1]. In preparation for
being able to run tests with `--no-default-features` add the "std"
feature as a requirement for all three examples. While we are at it use
the correct rand feature requirement: `rand-std`.
[1] Technically we only need "alloc" but "alloc" is not working with
Rust 1.29 currently so just use "std".
The `Debug` implementation for secrets is feature gated on `std` because
it uses a hasher from `std`. If `bitcoin_hashes` is enabled we can use
it for hashing. If neither `std` nor `bitcoin_hashes` is enabled fall
back to outputting:
<secret requires std or bitcoin_hashes feature to display>
Remove the docs conditional since we now implement `Debug` always.
When building with --no-default-features the compiler emits:
warning: unused import: `mem`
The call site is feature gated so we either need to feature gate the
import or use a fully qualified path. Since 'core' is quite short elect
to use the fully qualified path.
aa828f01a5 Improve documentation in the key module (Tobin Harding)
9e46d6f122 Add examples to types and methods in key module (Tobin Harding)
a7f3d9bcfd Improve key module docs (Tobin Harding)
6d23614467 Improve lib.rs rustdocs (Tobin Harding)
4c4268f1ad Improve docs on method generate_keypair (Tobin Harding)
Pull request description:
This PR is an initial attempt to more thoroughly test our public API.
Add examples to various types/methods/functions in the key module.
I'm not entirely sure when is enough, do we want an example on every single public method, function, and type or is this overkill. In this PR I tried to find a balance by doing ever method/function that took an argument that is a custom type from this lib. I think this should be extended to include return values too though ...
Thanks to @thomaseizinger for the idea!
First 2 patches are docs improvements to `lib.rs`.
ACKs for top commit:
apoelstra:
ACK aa828f01a5
Tree-SHA512: 9383ad263469f98ce7e988d47edc1482a09a0ce82f43d3991bd80aabdf621430f4a3c86be4debf33232dcb1d60d3e81f2c6d930ea7de7aa0e34b037accd7bc98
6fad20ef0c Fix the mess around Parity (Tobin Harding)
Pull request description:
Recently we made a wee mess with the `Parity` opaque type. Let's fix it
up by doing:
- Use an enum with variants `Even` and `Odd`.
- Add explicit conversion methods to/from u8 and i32
- Implement `BitXor`
This PR attempts to follow the plan laid out in the issue but:
- I don't get why`to_u8` is requested, I added it anyways.
- `to_i32` is not mentioned but we need it if we are going to pass the parity back to FFI code after receiving it _without_ having to care about the value. And that is the whole point of this `Parity` type in the first place.
- I don't get why `from_xxx` is fallible, when is an integer not even or odd?
Please note: This patch is an API breaking change that does _not_ follow the deprecation guidelines. Rust does not allow deprecating `From` impl blocks AFAICT.
Fixes: #370
ACKs for top commit:
apoelstra:
ACK 6fad20ef0c
Tree-SHA512: 810d5028f02fa608eab48721d32dca58be472080fcbac7a0ee79b5211b229112a756c48233e8597fb6c137690b2565431f410e9e97d6940ed389a4501bb015a0
We recently patched much of the docs in the `key` module, lets attempt
to attain perfection.
Improve docs by doing:
- Use full stops
- Use 100 character column width
- Use plural third person tense
- Use plural for section headings
- Fix any grammar mistakes
- Use code ticks and links as appropriate
Done in an effort to better test our public API.
Add tests in the `Examples` section as is idiomatic in the Rust
ecosystem.
Make other minor improvements to any rusdocs we touch:
- Use full stops
- Use 100 character column width
- Use plural third person tense
- Use plural for section headings
Improve method docs by doing:
- Remove 'batch' comment
- Remove mention of required features, docs already show this
- Use links to types as well as ticks
Recently we made a wee mess with the `Parity` opaque type. Let's fix it
up by doing:
- Use an enum with variants `Even` and `Odd`.
- Add explicit conversion methods to/from u8 and i32
- Implement `BitXor`
Note: This patch is an API breaking change that does _not_ follow the
deprecation guidelines. Rust does not allow deprecating `From` impl
blocks AFAICT.
20fe3a14dc Add a disabled rustfmt.toml (Tobin Harding)
Pull request description:
We do not currently use `rustfmt`, this is a nuisance for devs who routinely work on code bases that do use rustfmt who often have their editors set to format on save. We can make the life of such devs much better by explicitly disabling formatting using `rustfmt.toml`.
ref: https://rust-lang.github.io/rustfmt/?version=v1.4.38&search=#disable_all_formatting
ACKs for top commit:
apoelstra:
ACK 20fe3a14dc
Tree-SHA512: 63a7f9c64886cf85e15748c28137a9e642de74890c0b5194fafe3b6c0f9243662b9b7c19c04fbf56e9dd4b14449b07cefd163f69ccfa7e4ffe7dbc3df60b34ff
Andrew Poelstra
Martin Habovstiak
Tobin Harding
Vincent Liao