751 lines
24 KiB
Rust
751 lines
24 KiB
Rust
// SPDX-License-Identifier: CC0-1.0
|
|
|
|
//! Bitcoin scripts.
|
|
//!
|
|
//! *[See also the `Script` type](Script).*
|
|
//!
|
|
//! This module provides the structures and functions needed to support scripts.
|
|
//!
|
|
//! <details>
|
|
//! <summary>What is Bitcoin script</summary>
|
|
//!
|
|
//! Scripts define Bitcoin's digital signature scheme: a signature is formed
|
|
//! from a script (the second half of which is defined by a coin to be spent,
|
|
//! and the first half provided by the spending transaction), and is valid iff
|
|
//! the script leaves `TRUE` on the stack after being evaluated. Bitcoin's
|
|
//! script is a stack-based assembly language similar in spirit to [Forth].
|
|
//!
|
|
//! Script is represented as a sequence of bytes on the wire, each byte representing an operation,
|
|
//! or data to be pushed on the stack.
|
|
//!
|
|
//! See [Bitcoin Wiki: Script][wiki-script] for more information.
|
|
//!
|
|
//! [Forth]: https://en.wikipedia.org/wiki/Forth_(programming_language)
|
|
//!
|
|
//! [wiki-script]: https://en.bitcoin.it/wiki/Script
|
|
//! </details>
|
|
//!
|
|
//! In this library we chose to keep the byte representation in memory and decode opcodes only when
|
|
//! processing the script. This is similar to Rust choosing to represent strings as UTF-8-encoded
|
|
//! bytes rather than slice of `char`s. In both cases the individual items can have different sizes
|
|
//! and forcing them to be larger would waste memory and, in case of Bitcoin script, even some
|
|
//! performance (forcing allocations).
|
|
//!
|
|
//! ## `Script` vs `ScriptBuf` vs `Builder`
|
|
//!
|
|
//! These are the most important types in this module and they are quite similar, so it may seem
|
|
//! confusing what the differences are. `Script` is an unsized type much like `str` or `Path` are
|
|
//! and `ScriptBuf` is an owned counterpart to `Script` just like `String` is an owned counterpart
|
|
//! to `str`.
|
|
//!
|
|
//! However it is common to construct an owned script and then pass it around. For this case a
|
|
//! builder API is more convenient. To support this we provide `Builder` type which is very similar
|
|
//! to `ScriptBuf` but its methods take `self` instead of `&mut self` and return `Self`. It also
|
|
//! contains a cache that may make some modifications faster. This cache is usually not needed
|
|
//! outside of creating the script.
|
|
//!
|
|
//! At the time of writing there's only one operation using the cache - `push_verify`, so the cache
|
|
//! is minimal but we may extend it in the future if needed.
|
|
|
|
mod borrowed;
|
|
mod builder;
|
|
mod instruction;
|
|
mod owned;
|
|
mod push_bytes;
|
|
#[cfg(test)]
|
|
mod tests;
|
|
pub mod witness_program;
|
|
pub mod witness_version;
|
|
|
|
use alloc::rc::Rc;
|
|
#[cfg(any(not(rust_v_1_60), target_has_atomic = "ptr"))]
|
|
use alloc::sync::Arc;
|
|
use core::borrow::{Borrow, BorrowMut};
|
|
use core::cmp::Ordering;
|
|
use core::fmt;
|
|
use core::ops::{Deref, DerefMut};
|
|
|
|
use hashes::{hash160, sha256};
|
|
use io::{BufRead, Write};
|
|
#[cfg(feature = "serde")]
|
|
use serde;
|
|
|
|
use crate::blockdata::opcodes::all::*;
|
|
use crate::blockdata::opcodes::{self, Opcode};
|
|
use crate::consensus::{encode, Decodable, Encodable};
|
|
use crate::internal_macros::impl_asref_push_bytes;
|
|
use crate::prelude::*;
|
|
use crate::{io, OutPoint};
|
|
|
|
#[rustfmt::skip] // Keep public re-exports separate.
|
|
#[doc(inline)]
|
|
pub use self::{
|
|
borrowed::*,
|
|
builder::*,
|
|
instruction::*,
|
|
owned::*,
|
|
push_bytes::*,
|
|
};
|
|
|
|
hashes::hash_newtype! {
|
|
/// A hash of Bitcoin Script bytecode.
|
|
pub struct ScriptHash(hash160::Hash);
|
|
/// SegWit version of a Bitcoin Script bytecode hash.
|
|
pub struct WScriptHash(sha256::Hash);
|
|
}
|
|
impl_asref_push_bytes!(ScriptHash, WScriptHash);
|
|
|
|
impl From<ScriptBuf> for ScriptHash {
|
|
fn from(script: ScriptBuf) -> ScriptHash { script.script_hash() }
|
|
}
|
|
|
|
impl From<&ScriptBuf> for ScriptHash {
|
|
fn from(script: &ScriptBuf) -> ScriptHash { script.script_hash() }
|
|
}
|
|
|
|
impl From<&Script> for ScriptHash {
|
|
fn from(script: &Script) -> ScriptHash { script.script_hash() }
|
|
}
|
|
|
|
impl From<ScriptBuf> for WScriptHash {
|
|
fn from(script: ScriptBuf) -> WScriptHash { script.wscript_hash() }
|
|
}
|
|
|
|
impl From<&ScriptBuf> for WScriptHash {
|
|
fn from(script: &ScriptBuf) -> WScriptHash { script.wscript_hash() }
|
|
}
|
|
|
|
impl From<&Script> for WScriptHash {
|
|
fn from(script: &Script) -> WScriptHash { script.wscript_hash() }
|
|
}
|
|
|
|
/// Encodes an integer in script(minimal CScriptNum) format.
|
|
///
|
|
/// Writes bytes into the buffer and returns the number of bytes written.
|
|
///
|
|
/// Note that `write_scriptint`/`read_scriptint` do not roundtrip if the value written requires
|
|
/// more than 4 bytes, this is in line with Bitcoin Core (see [`CScriptNum::serialize`]).
|
|
///
|
|
/// [`CScriptNum::serialize`]: <https://github.com/bitcoin/bitcoin/blob/8ae2808a4354e8dcc697f76bacc5e2f2befe9220/src/script/script.h#L345>
|
|
pub fn write_scriptint(out: &mut [u8; 8], n: i64) -> usize {
|
|
let mut len = 0;
|
|
if n == 0 {
|
|
return len;
|
|
}
|
|
|
|
let neg = n < 0;
|
|
|
|
let mut abs = n.unsigned_abs();
|
|
while abs > 0xFF {
|
|
out[len] = (abs & 0xFF) as u8;
|
|
len += 1;
|
|
abs >>= 8;
|
|
}
|
|
// If the number's value causes the sign bit to be set, we need an extra
|
|
// byte to get the correct value and correct sign bit
|
|
if abs & 0x80 != 0 {
|
|
out[len] = abs as u8;
|
|
len += 1;
|
|
out[len] = if neg { 0x80u8 } else { 0u8 };
|
|
len += 1;
|
|
}
|
|
// Otherwise we just set the sign bit ourselves
|
|
else {
|
|
abs |= if neg { 0x80 } else { 0 };
|
|
out[len] = abs as u8;
|
|
len += 1;
|
|
}
|
|
len
|
|
}
|
|
|
|
/// Decodes an integer in script(minimal CScriptNum) format.
|
|
///
|
|
/// Notice that this fails on overflow: the result is the same as in
|
|
/// bitcoind, that only 4-byte signed-magnitude values may be read as
|
|
/// numbers. They can be added or subtracted (and a long time ago,
|
|
/// multiplied and divided), and this may result in numbers which
|
|
/// can't be written out in 4 bytes or less. This is ok! The number
|
|
/// just can't be read as a number again.
|
|
/// This is a bit crazy and subtle, but it makes sense: you can load
|
|
/// 32-bit numbers and do anything with them, which back when mult/div
|
|
/// was allowed, could result in up to a 64-bit number. We don't want
|
|
/// overflow since that's surprising --- and we don't want numbers that
|
|
/// don't fit in 64 bits (for efficiency on modern processors) so we
|
|
/// simply say, anything in excess of 32 bits is no longer a number.
|
|
/// This is basically a ranged type implementation.
|
|
///
|
|
/// This code is based on the `CScriptNum` constructor in Bitcoin Core (see `script.h`).
|
|
pub fn read_scriptint(v: &[u8]) -> Result<i64, Error> {
|
|
let last = match v.last() {
|
|
Some(last) => last,
|
|
None => return Ok(0),
|
|
};
|
|
if v.len() > 4 {
|
|
return Err(Error::NumericOverflow);
|
|
}
|
|
// Comment and code copied from Bitcoin Core:
|
|
// https://github.com/bitcoin/bitcoin/blob/447f50e4aed9a8b1d80e1891cda85801aeb80b4e/src/script/script.h#L247-L262
|
|
// If the most-significant-byte - excluding the sign bit - is zero
|
|
// then we're not minimal. Note how this test also rejects the
|
|
// negative-zero encoding, 0x80.
|
|
if (*last & 0x7f) == 0 {
|
|
// One exception: if there's more than one byte and the most
|
|
// significant bit of the second-most-significant-byte is set
|
|
// it would conflict with the sign bit. An example of this case
|
|
// is +-255, which encode to 0xff00 and 0xff80 respectively.
|
|
// (big-endian).
|
|
if v.len() <= 1 || (v[v.len() - 2] & 0x80) == 0 {
|
|
return Err(Error::NonMinimalPush);
|
|
}
|
|
}
|
|
|
|
Ok(scriptint_parse(v))
|
|
}
|
|
|
|
/// Decodes an integer in script format without non-minimal error.
|
|
///
|
|
/// The overflow error for slices over 4 bytes long is still there.
|
|
/// See [`read_scriptint`] for a description of some subtleties of
|
|
/// this function.
|
|
pub fn read_scriptint_non_minimal(v: &[u8]) -> Result<i64, Error> {
|
|
if v.is_empty() {
|
|
return Ok(0);
|
|
}
|
|
if v.len() > 4 {
|
|
return Err(Error::NumericOverflow);
|
|
}
|
|
|
|
Ok(scriptint_parse(v))
|
|
}
|
|
|
|
// Caller to guarantee that `v` is not empty.
|
|
fn scriptint_parse(v: &[u8]) -> i64 {
|
|
let (mut ret, sh) = v.iter().fold((0, 0), |(acc, sh), n| (acc + ((*n as i64) << sh), sh + 8));
|
|
if v[v.len() - 1] & 0x80 != 0 {
|
|
ret &= (1 << (sh - 1)) - 1;
|
|
ret = -ret;
|
|
}
|
|
ret
|
|
}
|
|
|
|
/// Decodes a boolean.
|
|
///
|
|
/// This is like "`read_scriptint` then map 0 to false and everything
|
|
/// else as true", except that the overflow rules don't apply.
|
|
#[inline]
|
|
pub fn read_scriptbool(v: &[u8]) -> bool {
|
|
match v.split_last() {
|
|
Some((last, rest)) => !((last & !0x80 == 0x00) && rest.iter().all(|&b| b == 0)),
|
|
None => false,
|
|
}
|
|
}
|
|
|
|
// We internally use implementation based on iterator so that it automatically advances as needed
|
|
// Errors are same as above, just different type.
|
|
fn read_uint_iter(data: &mut core::slice::Iter<'_, u8>, size: usize) -> Result<usize, UintError> {
|
|
if data.len() < size {
|
|
Err(UintError::EarlyEndOfScript)
|
|
} else if size > usize::from(u16::MAX / 8) {
|
|
// Casting to u32 would overflow
|
|
Err(UintError::NumericOverflow)
|
|
} else {
|
|
let mut ret = 0;
|
|
for (i, item) in data.take(size).enumerate() {
|
|
ret = usize::from(*item)
|
|
// Casting is safe because we checked above to not repeat the same check in a loop
|
|
.checked_shl((i * 8) as u32)
|
|
.ok_or(UintError::NumericOverflow)?
|
|
.checked_add(ret)
|
|
.ok_or(UintError::NumericOverflow)?;
|
|
}
|
|
Ok(ret)
|
|
}
|
|
}
|
|
|
|
fn opcode_to_verify(opcode: Option<Opcode>) -> Option<Opcode> {
|
|
opcode.and_then(|opcode| match opcode {
|
|
OP_EQUAL => Some(OP_EQUALVERIFY),
|
|
OP_NUMEQUAL => Some(OP_NUMEQUALVERIFY),
|
|
OP_CHECKSIG => Some(OP_CHECKSIGVERIFY),
|
|
OP_CHECKMULTISIG => Some(OP_CHECKMULTISIGVERIFY),
|
|
_ => None,
|
|
})
|
|
}
|
|
|
|
// We keep all the `Script` and `ScriptBuf` impls together since its easier to see side-by-side.
|
|
|
|
impl From<ScriptBuf> for Box<Script> {
|
|
fn from(v: ScriptBuf) -> Self { v.into_boxed_script() }
|
|
}
|
|
|
|
impl From<ScriptBuf> for Cow<'_, Script> {
|
|
fn from(value: ScriptBuf) -> Self { Cow::Owned(value) }
|
|
}
|
|
|
|
impl<'a> From<Cow<'a, Script>> for ScriptBuf {
|
|
fn from(value: Cow<'a, Script>) -> Self {
|
|
match value {
|
|
Cow::Owned(owned) => owned,
|
|
Cow::Borrowed(borrwed) => borrwed.into(),
|
|
}
|
|
}
|
|
}
|
|
|
|
impl<'a> From<Cow<'a, Script>> for Box<Script> {
|
|
fn from(value: Cow<'a, Script>) -> Self {
|
|
match value {
|
|
Cow::Owned(owned) => owned.into(),
|
|
Cow::Borrowed(borrwed) => borrwed.into(),
|
|
}
|
|
}
|
|
}
|
|
|
|
impl<'a> From<&'a Script> for Box<Script> {
|
|
fn from(value: &'a Script) -> Self { value.to_owned().into() }
|
|
}
|
|
|
|
impl<'a> From<&'a Script> for ScriptBuf {
|
|
fn from(value: &'a Script) -> Self { value.to_owned() }
|
|
}
|
|
|
|
impl<'a> From<&'a Script> for Cow<'a, Script> {
|
|
fn from(value: &'a Script) -> Self { Cow::Borrowed(value) }
|
|
}
|
|
|
|
/// Note: This will fail to compile on old Rust for targets that don't support atomics
|
|
#[cfg(any(not(rust_v_1_60), target_has_atomic = "ptr"))]
|
|
impl<'a> From<&'a Script> for Arc<Script> {
|
|
fn from(value: &'a Script) -> Self {
|
|
let rw: *const [u8] = Arc::into_raw(Arc::from(&value.0));
|
|
// SAFETY: copied from `std`
|
|
// The pointer was just created from an Arc without deallocating
|
|
// Casting a slice to a transparent struct wrapping that slice is sound (same
|
|
// layout).
|
|
unsafe { Arc::from_raw(rw as *const Script) }
|
|
}
|
|
}
|
|
|
|
impl<'a> From<&'a Script> for Rc<Script> {
|
|
fn from(value: &'a Script) -> Self {
|
|
let rw: *const [u8] = Rc::into_raw(Rc::from(&value.0));
|
|
// SAFETY: copied from `std`
|
|
// The pointer was just created from an Rc without deallocating
|
|
// Casting a slice to a transparent struct wrapping that slice is sound (same
|
|
// layout).
|
|
unsafe { Rc::from_raw(rw as *const Script) }
|
|
}
|
|
}
|
|
|
|
impl From<Vec<u8>> for ScriptBuf {
|
|
fn from(v: Vec<u8>) -> Self { ScriptBuf(v) }
|
|
}
|
|
|
|
impl From<ScriptBuf> for Vec<u8> {
|
|
fn from(v: ScriptBuf) -> Self { v.0 }
|
|
}
|
|
|
|
impl AsRef<Script> for Script {
|
|
#[inline]
|
|
fn as_ref(&self) -> &Script { self }
|
|
}
|
|
|
|
impl AsRef<Script> for ScriptBuf {
|
|
fn as_ref(&self) -> &Script { self }
|
|
}
|
|
|
|
impl AsRef<[u8]> for Script {
|
|
#[inline]
|
|
fn as_ref(&self) -> &[u8] { self.as_bytes() }
|
|
}
|
|
|
|
impl AsRef<[u8]> for ScriptBuf {
|
|
fn as_ref(&self) -> &[u8] { self.as_bytes() }
|
|
}
|
|
|
|
impl AsMut<Script> for Script {
|
|
fn as_mut(&mut self) -> &mut Script { self }
|
|
}
|
|
|
|
impl AsMut<Script> for ScriptBuf {
|
|
fn as_mut(&mut self) -> &mut Script { self }
|
|
}
|
|
|
|
impl AsMut<[u8]> for Script {
|
|
#[inline]
|
|
fn as_mut(&mut self) -> &mut [u8] { self.as_mut_bytes() }
|
|
}
|
|
|
|
impl AsMut<[u8]> for ScriptBuf {
|
|
fn as_mut(&mut self) -> &mut [u8] { self.as_mut_bytes() }
|
|
}
|
|
|
|
impl fmt::Debug for Script {
|
|
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
|
|
f.write_str("Script(")?;
|
|
self.fmt_asm(f)?;
|
|
f.write_str(")")
|
|
}
|
|
}
|
|
|
|
impl fmt::Debug for ScriptBuf {
|
|
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { fmt::Debug::fmt(self.as_script(), f) }
|
|
}
|
|
|
|
impl fmt::Display for Script {
|
|
#[inline]
|
|
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { self.fmt_asm(f) }
|
|
}
|
|
|
|
impl fmt::Display for ScriptBuf {
|
|
#[inline]
|
|
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { fmt::Display::fmt(self.as_script(), f) }
|
|
}
|
|
|
|
impl fmt::LowerHex for Script {
|
|
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
|
|
fmt::LowerHex::fmt(&self.as_bytes().as_hex(), f)
|
|
}
|
|
}
|
|
|
|
impl fmt::LowerHex for ScriptBuf {
|
|
#[inline]
|
|
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { fmt::LowerHex::fmt(self.as_script(), f) }
|
|
}
|
|
|
|
impl fmt::UpperHex for Script {
|
|
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
|
|
fmt::UpperHex::fmt(&self.as_bytes().as_hex(), f)
|
|
}
|
|
}
|
|
|
|
impl fmt::UpperHex for ScriptBuf {
|
|
#[inline]
|
|
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result { fmt::UpperHex::fmt(self.as_script(), f) }
|
|
}
|
|
|
|
impl Deref for ScriptBuf {
|
|
type Target = Script;
|
|
|
|
fn deref(&self) -> &Self::Target { Script::from_bytes(&self.0) }
|
|
}
|
|
|
|
impl DerefMut for ScriptBuf {
|
|
fn deref_mut(&mut self) -> &mut Self::Target { Script::from_bytes_mut(&mut self.0) }
|
|
}
|
|
|
|
impl Borrow<Script> for ScriptBuf {
|
|
fn borrow(&self) -> &Script { self }
|
|
}
|
|
|
|
impl BorrowMut<Script> for ScriptBuf {
|
|
fn borrow_mut(&mut self) -> &mut Script { self }
|
|
}
|
|
|
|
impl PartialEq<ScriptBuf> for Script {
|
|
fn eq(&self, other: &ScriptBuf) -> bool { self.eq(other.as_script()) }
|
|
}
|
|
|
|
impl PartialEq<Script> for ScriptBuf {
|
|
fn eq(&self, other: &Script) -> bool { self.as_script().eq(other) }
|
|
}
|
|
|
|
impl PartialOrd<Script> for ScriptBuf {
|
|
fn partial_cmp(&self, other: &Script) -> Option<Ordering> {
|
|
self.as_script().partial_cmp(other)
|
|
}
|
|
}
|
|
|
|
impl PartialOrd<ScriptBuf> for Script {
|
|
fn partial_cmp(&self, other: &ScriptBuf) -> Option<Ordering> {
|
|
self.partial_cmp(other.as_script())
|
|
}
|
|
}
|
|
|
|
#[cfg(feature = "serde")]
|
|
impl serde::Serialize for Script {
|
|
/// User-facing serialization for `Script`.
|
|
fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>
|
|
where
|
|
S: serde::Serializer,
|
|
{
|
|
if serializer.is_human_readable() {
|
|
serializer.collect_str(&format_args!("{:x}", self))
|
|
} else {
|
|
serializer.serialize_bytes(self.as_bytes())
|
|
}
|
|
}
|
|
}
|
|
|
|
/// Can only deserialize borrowed bytes.
|
|
#[cfg(feature = "serde")]
|
|
impl<'de> serde::Deserialize<'de> for &'de Script {
|
|
fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>
|
|
where
|
|
D: serde::Deserializer<'de>,
|
|
{
|
|
if deserializer.is_human_readable() {
|
|
use crate::serde::de::Error;
|
|
|
|
return Err(D::Error::custom(
|
|
"deserialization of `&Script` from human-readable formats is not possible",
|
|
));
|
|
}
|
|
|
|
struct Visitor;
|
|
impl<'de> serde::de::Visitor<'de> for Visitor {
|
|
type Value = &'de Script;
|
|
|
|
fn expecting(&self, formatter: &mut fmt::Formatter) -> fmt::Result {
|
|
formatter.write_str("borrowed bytes")
|
|
}
|
|
|
|
fn visit_borrowed_bytes<E>(self, v: &'de [u8]) -> Result<Self::Value, E>
|
|
where
|
|
E: serde::de::Error,
|
|
{
|
|
Ok(Script::from_bytes(v))
|
|
}
|
|
}
|
|
deserializer.deserialize_bytes(Visitor)
|
|
}
|
|
}
|
|
|
|
#[cfg(feature = "serde")]
|
|
impl serde::Serialize for ScriptBuf {
|
|
/// User-facing serialization for `Script`.
|
|
fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error>
|
|
where
|
|
S: serde::Serializer,
|
|
{
|
|
(**self).serialize(serializer)
|
|
}
|
|
}
|
|
|
|
#[cfg(feature = "serde")]
|
|
impl<'de> serde::Deserialize<'de> for ScriptBuf {
|
|
fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>
|
|
where
|
|
D: serde::Deserializer<'de>,
|
|
{
|
|
use core::fmt::Formatter;
|
|
|
|
use hex::FromHex;
|
|
|
|
if deserializer.is_human_readable() {
|
|
struct Visitor;
|
|
impl<'de> serde::de::Visitor<'de> for Visitor {
|
|
type Value = ScriptBuf;
|
|
|
|
fn expecting(&self, formatter: &mut Formatter) -> fmt::Result {
|
|
formatter.write_str("a script hex")
|
|
}
|
|
|
|
fn visit_str<E>(self, v: &str) -> Result<Self::Value, E>
|
|
where
|
|
E: serde::de::Error,
|
|
{
|
|
let v = Vec::from_hex(v).map_err(E::custom)?;
|
|
Ok(ScriptBuf::from(v))
|
|
}
|
|
}
|
|
deserializer.deserialize_str(Visitor)
|
|
} else {
|
|
struct BytesVisitor;
|
|
|
|
impl<'de> serde::de::Visitor<'de> for BytesVisitor {
|
|
type Value = ScriptBuf;
|
|
|
|
fn expecting(&self, formatter: &mut Formatter) -> fmt::Result {
|
|
formatter.write_str("a script Vec<u8>")
|
|
}
|
|
|
|
fn visit_bytes<E>(self, v: &[u8]) -> Result<Self::Value, E>
|
|
where
|
|
E: serde::de::Error,
|
|
{
|
|
Ok(ScriptBuf::from(v.to_vec()))
|
|
}
|
|
|
|
fn visit_byte_buf<E>(self, v: Vec<u8>) -> Result<Self::Value, E>
|
|
where
|
|
E: serde::de::Error,
|
|
{
|
|
Ok(ScriptBuf::from(v))
|
|
}
|
|
}
|
|
deserializer.deserialize_byte_buf(BytesVisitor)
|
|
}
|
|
}
|
|
}
|
|
|
|
impl Encodable for Script {
|
|
#[inline]
|
|
fn consensus_encode<W: Write + ?Sized>(&self, w: &mut W) -> Result<usize, io::Error> {
|
|
crate::consensus::encode::consensus_encode_with_size(&self.0, w)
|
|
}
|
|
}
|
|
|
|
impl Encodable for ScriptBuf {
|
|
#[inline]
|
|
fn consensus_encode<W: Write + ?Sized>(&self, w: &mut W) -> Result<usize, io::Error> {
|
|
self.0.consensus_encode(w)
|
|
}
|
|
}
|
|
|
|
impl Decodable for ScriptBuf {
|
|
#[inline]
|
|
fn consensus_decode_from_finite_reader<R: BufRead + ?Sized>(
|
|
r: &mut R,
|
|
) -> Result<Self, encode::Error> {
|
|
Ok(ScriptBuf(Decodable::consensus_decode_from_finite_reader(r)?))
|
|
}
|
|
}
|
|
|
|
/// Writes the assembly decoding of the script bytes to the formatter.
|
|
pub(super) fn bytes_to_asm_fmt(script: &[u8], f: &mut dyn fmt::Write) -> fmt::Result {
|
|
// This has to be a macro because it needs to break the loop
|
|
macro_rules! read_push_data_len {
|
|
($iter:expr, $len:literal, $formatter:expr) => {
|
|
match read_uint_iter($iter, $len) {
|
|
Ok(n) => {
|
|
n
|
|
},
|
|
Err(UintError::EarlyEndOfScript) => {
|
|
$formatter.write_str("<unexpected end>")?;
|
|
break;
|
|
}
|
|
// We got the data in a slice which implies it being shorter than `usize::MAX`
|
|
// So if we got overflow, we can confidently say the number is higher than length of
|
|
// the slice even though we don't know the exact number. This implies attempt to push
|
|
// past end.
|
|
Err(UintError::NumericOverflow) => {
|
|
$formatter.write_str("<push past end>")?;
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
let mut iter = script.iter();
|
|
// Was at least one opcode emitted?
|
|
let mut at_least_one = false;
|
|
// `iter` needs to be borrowed in `read_push_data_len`, so we have to use `while let` instead
|
|
// of `for`.
|
|
while let Some(byte) = iter.next() {
|
|
let opcode = Opcode::from(*byte);
|
|
|
|
let data_len = if let opcodes::Class::PushBytes(n) =
|
|
opcode.classify(opcodes::ClassifyContext::Legacy)
|
|
{
|
|
n as usize
|
|
} else {
|
|
match opcode {
|
|
OP_PUSHDATA1 => {
|
|
// side effects: may write and break from the loop
|
|
read_push_data_len!(&mut iter, 1, f)
|
|
}
|
|
OP_PUSHDATA2 => {
|
|
// side effects: may write and break from the loop
|
|
read_push_data_len!(&mut iter, 2, f)
|
|
}
|
|
OP_PUSHDATA4 => {
|
|
// side effects: may write and break from the loop
|
|
read_push_data_len!(&mut iter, 4, f)
|
|
}
|
|
_ => 0,
|
|
}
|
|
};
|
|
|
|
if at_least_one {
|
|
f.write_str(" ")?;
|
|
} else {
|
|
at_least_one = true;
|
|
}
|
|
// Write the opcode
|
|
if opcode == OP_PUSHBYTES_0 {
|
|
f.write_str("OP_0")?;
|
|
} else {
|
|
write!(f, "{:?}", opcode)?;
|
|
}
|
|
// Write any pushdata
|
|
if data_len > 0 {
|
|
f.write_str(" ")?;
|
|
if data_len <= iter.len() {
|
|
for ch in iter.by_ref().take(data_len) {
|
|
write!(f, "{:02x}", ch)?;
|
|
}
|
|
} else {
|
|
f.write_str("<push past end>")?;
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
Ok(())
|
|
}
|
|
|
|
/// Ways that a script might fail. Not everything is split up as
|
|
/// much as it could be; patches welcome if more detailed errors
|
|
/// would help you.
|
|
#[derive(Debug, Clone, PartialEq, Eq)]
|
|
#[non_exhaustive]
|
|
pub enum Error {
|
|
/// Something did a non-minimal push; for more information see
|
|
/// <https://github.com/bitcoin/bips/blob/master/bip-0062.mediawiki#push-operators>
|
|
NonMinimalPush,
|
|
/// Some opcode expected a parameter but it was missing or truncated.
|
|
EarlyEndOfScript,
|
|
/// Tried to read an array off the stack as a number when it was more than 4 bytes.
|
|
NumericOverflow,
|
|
/// Can not find the spent output.
|
|
UnknownSpentOutput(OutPoint),
|
|
/// Can not serialize the spending transaction.
|
|
Serialization,
|
|
}
|
|
|
|
impl fmt::Display for Error {
|
|
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
|
|
use Error::*;
|
|
|
|
match *self {
|
|
NonMinimalPush => f.write_str("non-minimal datapush"),
|
|
EarlyEndOfScript => f.write_str("unexpected end of script"),
|
|
NumericOverflow =>
|
|
f.write_str("numeric overflow (number on stack larger than 4 bytes)"),
|
|
UnknownSpentOutput(ref point) => write!(f, "unknown spent output: {}", point),
|
|
Serialization =>
|
|
f.write_str("can not serialize the spending transaction in Transaction::verify()"),
|
|
}
|
|
}
|
|
}
|
|
|
|
#[cfg(feature = "std")]
|
|
impl std::error::Error for Error {
|
|
fn source(&self) -> Option<&(dyn std::error::Error + 'static)> {
|
|
use Error::*;
|
|
|
|
match *self {
|
|
NonMinimalPush
|
|
| EarlyEndOfScript
|
|
| NumericOverflow
|
|
| UnknownSpentOutput(_)
|
|
| Serialization => None,
|
|
}
|
|
}
|
|
}
|
|
|
|
// Our internal error proves that we only return these two cases from `read_uint_iter`.
|
|
// Since it's private we don't bother with trait impls besides From.
|
|
enum UintError {
|
|
EarlyEndOfScript,
|
|
NumericOverflow,
|
|
}
|
|
|
|
impl From<UintError> for Error {
|
|
fn from(error: UintError) -> Self {
|
|
match error {
|
|
UintError::EarlyEndOfScript => Error::EarlyEndOfScript,
|
|
UintError::NumericOverflow => Error::NumericOverflow,
|
|
}
|
|
}
|
|
}
|