feat: add annex to env

Author: stringhandler

fuzz/fuzz_lib/lib.rs

@@ -64,7 +64,7 @@ impl<'f> Extractor<'f> {
 
     /// Attempt to yield a type from the fuzzer.
     pub fn extract_final_type(&mut self) -> Option<Arc<FinalTy>> {
-        // We can costruct extremely large types by duplicating Arcs; there
+        // We can construct extremely large types by duplicating Arcs; there
         // is no need to have an exponential blowup in the number of tasks.
         const MAX_N_TASKS: usize = 300;
 
@@ -206,7 +206,7 @@ impl<'f> Extractor<'f> {
 
     /// Attempt to yield a type from the fuzzer.
     pub fn extract_old_final_type(&mut self) -> Option<Arc<OldFinalTy>> {
-        // We can costruct extremely large types by duplicating Arcs; there
+        // We can construct extremely large types by duplicating Arcs; there
         // is no need to have an exponential blowup in the number of tasks.
         const MAX_N_TASKS: usize = 300;
 

src/analysis.rs

@@ -6,6 +6,8 @@ use std::{cmp, fmt};
 use crate::value::Word;
 #[cfg(feature = "elements")]
 use elements::encode::Encodable;
+#[cfg(feature = "serde")]
+use serde::Serialize;
 #[cfg(feature = "elements")]
 use std::{convert::TryFrom, io};
 
@@ -65,6 +67,7 @@ impl From<U32Weight> for bitcoin::Weight {
 /// Programs that are CPU-heavy need to be padded
 /// so that the witness stack provides a large-enough budget.
 #[derive(Copy, Clone, Debug, Eq, PartialEq, Ord, PartialOrd, Hash)]
+#[cfg_attr(feature = "serde", derive(Serialize))]
 pub struct Cost(u32);
 
 impl Cost {
@@ -215,6 +218,7 @@ impl From<Cost> for bitcoin::Weight {
 
 /// Bounds on the resources required by a node during execution on the Bit Machine
 #[derive(Debug, Copy, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)]
+#[cfg_attr(feature = "serde", derive(Serialize))]
 pub struct NodeBounds {
     /// Upper bound on the required number of cells (bits).
     /// The root additionally requires the bit width of its source and target type (input, output)

src/dag.rs

@@ -347,7 +347,7 @@ pub trait DagLike: Sized {
 /// with the returned nodes. To correct this, you need to call
 /// [`PostOrderIterItem::unswap`].
 ///
-/// To avoid confusion, this structure cannot be directly costructed.
+/// To avoid confusion, this structure cannot be directly constructed.
 /// Instead it is implicit in the [`DagLike::rtl_post_order_iter`]
 /// method.
 #[derive(Clone, Debug)]

src/human_encoding/named_node.rs

@@ -504,7 +504,7 @@ pub struct Namer {
 }
 
 impl Namer {
-    /// Costruct a new `Namer`. Will assign the name `main` to the node with
+    /// Construct a new `Namer`. Will assign the name `main` to the node with
     /// the given CMR.
     pub fn new_rooted(root_cmr: Cmr) -> Self {
         Namer {
@@ -515,7 +515,7 @@ impl Namer {
         }
     }
 
-    /// Costruct a new `Namer`.
+    /// Construct a new `Namer`.
     pub fn new() -> Self {
         Namer {
             const_idx: 0,

src/jet/elements/c_env.rs

@@ -3,6 +3,7 @@
 //! High level APIs for creating C FFI compatible environment.
 //!
 
+use bitcoin::taproot::TAPROOT_ANNEX_PREFIX;
 use hashes::Hash;
 use std::os::raw::c_uchar;
 
@@ -33,7 +34,6 @@ struct RawOutputData {
 /// passed to the C FFI.
 #[derive(Debug)]
 struct RawInputData {
-    #[allow(dead_code)] // see FIXME below
     pub annex: Option<Vec<c_uchar>>,
     // pegin
     pub genesis_hash: Option<[c_uchar; 32]>,
@@ -75,8 +75,12 @@ fn new_raw_input<'raw>(
     inp_data: &'raw RawInputData,
 ) -> c_elements::CRawInput<'raw> {
     c_elements::CRawInput {
-        // FIXME actually pass the annex in; see https://github.com/BlockstreamResearch/simplicity/issues/311 for some difficulty here.
-        annex: core::ptr::null(),
+        annex: inp_data
+            .annex
+            .as_ref()
+            .map(|annex| c_elements::CRawBuffer::new(annex))
+            .as_ref()
+            .map_or(core::ptr::null(), |ptr| ptr as *const _),
         prev_txid: inp.previous_output.txid.as_ref(),
         pegin: inp_data.genesis_hash.as_ref(),
         issuance: if inp.has_issuance() {
@@ -114,7 +118,7 @@ fn new_tx_data(tx: &elements::Transaction, in_utxos: &[ElementsUtxo]) -> RawTran
     };
     for (inp, in_utxo) in tx.input.iter().zip(in_utxos.iter()) {
         let inp_data = RawInputData {
-            annex: None, // Actually store annex
+            annex: get_annex(&inp.witness).map(|s| s.to_vec()),
             genesis_hash: inp
                 .pegin_data()
                 .map(|x| x.genesis_hash.to_raw_hash().to_byte_array()),
@@ -148,14 +152,29 @@ pub(super) fn new_tx(
 ) -> *mut c_elements::CTransaction {
     let mut raw_inputs = Vec::new();
     let mut raw_outputs = Vec::new();
+    // Allocate space for the raw annexes. This dumb `Vec::from_iter` construction is
+    // equivalent to `vec![None; tx.input.len()]`, but that won't compile because it
+    // requires Option::<CRawBuffer>::None to be cloneable, which it's not because
+    // CRawBuffer isn't.
+
+    // SAFETY: this allocation *must* live until after the `simplicity_mallocTransaction`
+    //  at the bottom of this function. We convert the vector to a boxed slice to ensure
+    //  it cannot be resized, which would potentially trigger a reallocation.
+    let mut raw_annexes = Vec::from_iter((0..tx.input.len()).map(|_| None)).into_boxed_slice();
+
     let txid = tx.txid();
     let tx_data = new_tx_data(tx, in_utxos);
-    for ((inp, in_utxo), inp_data) in tx
+    for (((n, inp), in_utxo), inp_data) in tx
         .input
         .iter()
+        .enumerate()
         .zip(in_utxos.iter())
         .zip(tx_data.inputs.iter())
     {
+        raw_annexes[n] = inp_data
+            .annex
+            .as_ref()
+            .map(|annex| c_elements::CRawBuffer::new(annex));
         let res = new_raw_input(inp, in_utxo, inp_data);
         raw_inputs.push(res);
     }
@@ -172,10 +191,17 @@ pub(super) fn new_tx(
         version: tx.version,
         locktime: tx.lock_time.to_consensus_u32(),
     };
-    unsafe {
+    let ret = unsafe {
         // SAFETY: this is a FFI call and we constructed its argument correctly.
         c_elements::simplicity_mallocTransaction(&c_raw_tx)
-    }
+    };
+
+    // Explicitly drop raw_annexes so Rust doesn't try any funny business dropping it early.
+    // Drop raw_inputs first since it contains pointers into raw_annexes and we don't want
+    // them to dangle. (It'd be safe since they're raw pointers, but still bad mojo.)
+    drop(raw_inputs);
+    drop(raw_annexes);
+    ret
 }
 
 pub(super) fn new_tap_env(
@@ -256,3 +282,13 @@ fn serialize_surjection_proof(surjection_proof: &Option<Box<SurjectionProof>>) -
         .map(|x| x.serialize())
         .unwrap_or_default()
 }
+
+/// If the last item in the witness stack is an annex, return the data following the 0x50 byte.
+fn get_annex(in_witness: &elements::TxInWitness) -> Option<&[u8]> {
+    let last_item = in_witness.script_witness.last()?;
+    if *last_item.first()? == TAPROOT_ANNEX_PREFIX {
+        Some(&last_item[1..])
+    } else {
+        None
+    }
+}

src/node/convert.rs

@@ -58,7 +58,7 @@ pub enum Hide {
 /// 4. Finally, the node's data is passed to [`Self::convert_data`], whose job
 ///    it is to compute the cached data for the new node. For `case` combinators
 ///    where one child was pruned, `convert_data` will receive an `assertl` or
-///    `assertl`, as appropriate, rather than a `case`.
+///    `assertr`, as appropriate, rather than a `case`.
 ///
 /// If any method returns an error, then iteration is aborted immediately and
 /// the error returned to the caller. If the converter would like to recover

src/node/mod.rs

@@ -612,7 +612,7 @@ impl<N: Marker> Node<N> {
 
     /// Contruct a node from its constituent parts.
     ///
-    /// This method can be used to directly costruct a node. It will compute the CMR
+    /// This method can be used to directly construct a node. It will compute the CMR
     /// automatically based on the value of `inner` but requires that `cached_data`
     /// be provided.
     ///