So far we deserialized hex into `Vec<u8>` at run time. This was mainly
in tests where it had negligible performance cost. However moving the
computation to compile time has a few benefits: it allows proving the
length of the decoded bytes and identifies potential typos before the
code goes through LLVM and other compilation machinery which makes
feedback faster.
This change uses the `hex_lit` crate to move computation to compile
time. It is implemented as `const` declarative macro which doesn't blow
up compilation time.
In `bitcoin` when we use the `core2` dependency we always need the
"alloc" feature. Enabling "alloc" when enabling "core2" in the "no-std"
feature is confusing because it makes it seem that we don't always need
it.
Set usage of the "alloc" feature of `core2` in the `features` list.
`core2` is for Read/Write, nothing to do with allocation and we do not
use the "alloc" feature of `core2` in `hashes`.
Fix core2 dependency/features by doing:
- Explicitly enable "bitcoin_hashes/core2" in `bitcoin`.
- Do not enable "core2/alloc" in `hashes`
Improve all manifest package sections by doing:
- Order the list of options uniformly
- Remove unnecessary homepage option (currently same as repo)
- Add categories section
We can check which files are included in the packaged release with
`cargo package --list `.
Add an `exclude` section to each manifest that excludes `tests/` and
`contrib/`. Not all crates have a `tests/` directory yet but they should
so add the exclude anyway to future proof the crates.
The rust version is supposed to be set by the build script so that users
automagically get features matching the toolchain in use. Currently we
have a feature in the manifest for `rustv_1_53` instead setting a
compiler conditional configuration option in the build script. This
causes `cargo +1.41.1 --all-features check` to fail.
The `bip158` module uses a `HashSet` and in order to do so requires the
`hashbrown` dependency for "no-std" builds.
We can replace the usage of `HashSet` with a `BTreeSet` in `bip158` and
remove the `hashbrown` dependency entirely.
This patch makes no claims about performance cost or benefit of this
change. The patch also makes no claims about the validity of the current
`HashSet` usage.
The `hashbrown` dependency and `HashSet` usage can be trivially added
back in if someone comes up with perf data to back it up.
941083ec4e Remove rand-std dev-dependency from secp256k1 (Tobin C. Harding)
f71335f971 Add rand-std feature (Tobin C. Harding)
Pull request description:
This PR uncovered incorrect feature gating in `secp256k1`, fixed in https://github.com/rust-bitcoin/rust-secp256k1/pull/519
Currently we enable "secp256k1/rand-std" in the "rand" feature, this is incorrect because it means "rand" implies "std" which it does not.
Add a "rand-std" feature that turns on "seck256k1/rand-std" and make the "rand" feature turn on "seck256k1/rand".
Fix: #1384
ACKs for top commit:
sanket1729:
ACK 941083ec4e
apoelstra:
ACK 941083ec4e
Tree-SHA512: e1d12196766c2b3082b488fe7d0c03a865ebbfc70ffa6f128c57074df14da71e56c31ea92982991d376ac62fa86c6639049ce17aac93613b523ddcc99677c269
8ce928b8e7 Add testing section to readme (Tobin C. Harding)
2e79a0bdc4 Introduce mutation testing (Tobin C. Harding)
Pull request description:
Introduce mutation testing by way of mutagen [0] (see #1484 for context).
- Conditionally add the dev-dependency `mutagen` (using `do_mutate` flag)
This flag is not very well named but `mutagen` and `mutate` are already taken?
- Mutate all methods of the `U256` struct that do not require additional unit tests.
Uses `cfg(all(test, do_mutate), mutate)` - I cannot workout why we need to check on `test` as well i.e., I don't understand why we cannot use `cfg(do_mutate, mutate)`?
With this applied test can be run as usual with a stable toolchain. To run mutagen we use `RUSTFLAGS='--cfg=do_mutate' cargo +nightly mutagen` (doing so runs 29 mutants).
[0] https://github.com/llogiq/mutagen
ACKs for top commit:
Kixunil:
ACK 8ce928b8e7
apoelstra:
ACK 8ce928b8e7
Tree-SHA512: 024ba5d2dc983f7cd0444e09ba13280771157204999d2a44502e07a1d6436f571b75003c7cb543c943f17949b848d4070eda4e194bda774a3e41443ff79af0af
In order to get better test coverage we should not enable the secp26k1
feature "rand-std" in dev-dependencies but instead feature gate tests
that depend on this feature.
Currently we enable "secp256k1/rand-std" in the "rand" feature, this is
incorrect because it means "rand" implies "std" which it does not.
Add a "rand-std" feature that turns on "seck256k1/rand-std" and make the
"rand" feature turn on "seck256k1/rand".
Introduce mutation testing by way of mutagen [0]
- Conditionally add the dev-dependency `mutagen` (using `do_mutate`
flag)
This flag is not very well named but `mutagen` and `mutate` are already
taken?
- Mutate all methods of the `U256` struct that do not require additional
unit tests.
Uses `cfg(all(test, do_mutate), mutate)` - I cannot workout why we need
to check on `test` as well i.e., I don't understand why we cannot use
`cfg(do_mutate, mutate)`?
With this applied test can be run as usual with a stable toolchain. To
run mutagen we use `RUSTFLAGS='--cfg=do_mutate' cargo +nightly mutagen`.
[0] https://github.com/llogiq/mutagen
This renames `Script` to `ScriptBuf` and adds unsized `Script` modeled
after `PathBuf`/`Path`. The change cleans up the API a bit, especially
all functions that previously accepted `&Script` now accept truly
borrowed version. Some functions that perviously accepted `&[u8]` can
now accept `&Script` because constructing it is no loger costly.
In some protocols it is preferred to serialize consensus-encodable types
using consensus encoding. E.g. serialize `Transaction` as hex-encoded
string in Json in Bitcoin Core RPC protocol. This change provides
adapter to make this easier.
The adapter allows providing custom byte-to-string encoder for more
exotic cases and provides a hex implementation which should be useful in
majority of the cases.
Should help with #765
b6f9e47dba Fix `no_std` when `bitcoinconsensus` is enabled (Martin Habovstiak)
Pull request description:
`default-features = false` was missing previously but blindly adding it would lead to subtle risk of breaking when a crate not needing `std` depends on `bitcoinconsensus` and simultaneously another crate not needing `bitcoinconsensus` depends on `std` and another crate depends on them both.
This change fixes it by introducing `bitcoinconsensus-std` feature flag and provides a fallback if the flag is off. Unfortunately the fallback has to use a bit of reasonable `unsafe` due to limitations of upcasting.
The only safe alternatives are not do it and provide worse experience for crates that are affected by the problem above or break the API, which couldn't be backported and would be more annoying to use.
Closes#1343
This is considered PoC PR as I realized the possibility of the hack (and necessity of `unsafe`) at the last moment. Things like tests and modifying CONTRIBUTING to change the stance on `unsafe` will be added if `unsafe` is ACKed.
ACKs for top commit:
tcharding:
tACK b6f9e47dba
apoelstra:
ACK b6f9e47dba
Tree-SHA512: 3a2845f4701c94ff6214749fa490aecf3fd96089df31b15f9d3e0afe3c74329ff2b9054d51244358a79f928aa9d4cf4001fc3ec40a9b0e189323544c4480c709
1a89d5230c examples: Add taproot-psbt workflow example (Duncan Dean)
Pull request description:
Will address #893.
Currently includes a BIP86 example (no spendable script path)
Working on script path and key path spending when both are possible spending paths.
ACKs for top commit:
tcharding:
ACK 1a89d5230c
apoelstra:
ACK 1a89d5230c
Tree-SHA512: 31d23914faedb2632d517f9a827075f1bf387df6d98f151000f70546d1e67ac332e698c6a01bda40a9baf5b25bff0114928edc0879c5e01753ecfc6ad182fe26
`default-features = false` was missing previously but blindly adding it
would lead to subtle risk of breaking when a crate not needing `std`
depends on `bitcoinconsensus` and simultaneously another crate not
needing `bitcoinconsensus` depends on `std` and another crate depends on
them both.
This change fixes it by introducing `bitcoinconsensus-std` feature flag
and provides a fallback if the flag is off. Unfortunately the fallback
has to use a bit of reasonable `unsafe` due to limitations of upcasting.
The only safe alternatives are not do it and provide worse experience
for crates that are affected by the problem above or break the API,
which couldn't be backported and would be more annoying to use.
Closes#1343
This example shows how to use the PSBT API for taproot transactions.
We have a simple BIP86-style spend and an example of an inheritance
timelock that can be spent either by the beneficiary via the script
path after a timelock, or via the key path by the benefactor so that
they can refresh the timelock at any time.
`Cargo.toml` claimed that this crate works with very old versions of
`serde` which wasn't the case. This commit changes the versions to
minimal known-to-work values.
Add a new crate `bitcoin-internals` to be used for internal code needed
by multiple soon-to-be-created crates.
Add the `write_err` macro to `bitcoin-internals`, nothing else.
This patch uses a `path` dependency which means `rust-bitcoin` cannot be
released in its current state, will need to be changed once we release
the `bitcoin-internals` crate on `crates.io`.
Create a directory `bitcoin` and move into it the following as is with
no code changes:
- src
- Cargo.toml
- contrib
- test_data
- examples
Then do:
- Add a workspace to the repository root directory.
- Add the newly created `bitcoin` crate to the workspace.
- Exclude `fuzz` and `embedded` crates from the workspace.
- Add a contrib/test.sh script that runs contrib/test.sh in each
sub-crate
- Fix the bitcoin/contrib/test.sh script
Martin Habovstiak
Tobin C. Harding
Andrew Poelstra
Byron Hambly
Riccardo Casatta
Duncan Dean