ladybird/Libraries/LibJS/Rust/build.rs
Andreas Kling 4ac744082b LibJS: Cache dynamic environment coordinates
Dynamic environment binding opcodes lost the old coordinate warmup.
They were split away from the static coordinate opcodes. Hot closures
and eval-sensitive functions then resolved the same binding by name on
every execution, which regressed JS benchmark throughput badly.

Give each dynamic environment opcode a per-executable coordinate cache
slot. The cache keeps the bytecode stream immutable while letting both
interpreters take a direct declarative environment fast path after the
first lookup. Keep the existing eval invalidation behavior and only warm
caches for declarative-only chains so with environments continue to
observe object shadowing.

Reject cached bytecode that uses the no-cache sentinel for dynamic
environment coordinate cache operands, since execution indexes those
cache arrays unconditionally.

Rebaseline bytecode expectations for the instruction size changes. Add
coverage for with-object shadowing across repeated dynamic lookups and
for rejecting corrupt dynamic environment cache indices.
2026-05-19 15:54:23 +02:00

958 lines
36 KiB
Rust

/*
* Copyright (c) 2026-present, the Ladybird developers.
*
* SPDX-License-Identifier: BSD-2-Clause
*/
//! Build script that generates Rust bytecode instruction types from Bytecode.def.
//!
//! This mirrors Meta/Generators/generate_libjs_bytecode_def_derived.py but generates Rust
//! code instead of C++. The generated code lives in $OUT_DIR/instruction_generated.rs
//! and is included! from src/bytecode/instruction.rs.
use bytecode_def::{
Field, OpDef, STRUCT_ALIGN, compute_layouts, field_type_info, find_m_length_offset, round_up, user_fields,
};
use std::env;
use std::fs;
use std::io::Write;
use std::path::PathBuf;
fn rust_field_name(name: &str) -> String {
if let Some(stripped) = name.strip_prefix("m_") {
stripped.to_string()
} else {
name.to_string()
}
}
/// Find the count field corresponding to a given array field, using the same
/// heuristic as the Python generator: prefer `<name>_count`, fall back to
/// `<singularized name>_count`. Panics if no matching u32/size_t field is
/// found, mirroring the Python `find_count_field_name_or_die`.
fn find_count_field_name(op: &OpDef, array_field: &Field) -> String {
let mut candidates = vec![format!("{}_count", array_field.name)];
if let Some(stripped) = array_field.name.strip_suffix('s') {
candidates.push(format!("{stripped}_count"));
}
for candidate in &candidates {
for f in &op.fields {
if f.is_array {
continue;
}
if &f.name == candidate && (f.ty == "u32" || f.ty == "size_t") {
return candidate.clone();
}
}
}
panic!(
"No count field (u32/size_t) found for array field '{}' in op '{}'",
array_field.name, op.name
);
}
fn generate_rust_code(mut w: impl Write, ops: &[OpDef]) -> Result<(), Box<dyn std::error::Error>> {
writeln!(w, "// @generated from Libraries/LibJS/Bytecode/Bytecode.def")?;
writeln!(w, "// Do not edit manually.")?;
writeln!(w)?;
writeln!(w, "use super::operand::*;")?;
writeln!(w)?;
generate_opcode_enum(&mut w, ops)?;
generate_num_opcodes_const(&mut w, ops)?;
generate_instruction_enum(&mut w, ops)?;
generate_instruction_impl(&mut w, ops)?;
generate_instruction_length_from_bytes(&mut w, ops)?;
generate_validate_instruction(&mut w, ops)?;
Ok(())
}
fn generate_num_opcodes_const(mut w: impl Write, ops: &[OpDef]) -> Result<(), Box<dyn std::error::Error>> {
writeln!(w, "/// Number of distinct opcodes (the valid range for the type byte).")?;
writeln!(w, "pub const NUM_OPCODES: u32 = {};", ops.len())?;
writeln!(w)?;
Ok(())
}
fn generate_instruction_length_from_bytes(mut w: impl Write, ops: &[OpDef]) -> Result<(), Box<dyn std::error::Error>> {
writeln!(
w,
"/// Returns the encoded length in bytes of the instruction at `bytes[at..]`."
)?;
writeln!(
w,
"/// Reads `m_length` from the buffer for variable-length instructions; for fixed-"
)?;
writeln!(w, "/// length instructions, returns the statically-known size.")?;
writeln!(
w,
"pub fn instruction_length_from_bytes(opcode: u8, bytes: &[u8], at: usize) -> Result<usize, super::validator::ValidationErrorKind> {{"
)?;
writeln!(w, " use super::validator::ValidationErrorKind;")?;
writeln!(w, " match opcode {{")?;
for (i, op) in ops.iter().enumerate() {
let has_array = op.fields.iter().any(|f| f.is_array);
if !has_array {
let mut offset: usize = 2;
for f in &op.fields {
if f.is_array || f.name == "m_type" || f.name == "m_strict" {
continue;
}
let info = field_type_info(&f.ty);
offset = round_up(offset, info.align);
offset += info.size;
}
let final_size = round_up(offset, STRUCT_ALIGN);
let op_name = &op.name;
writeln!(w, " {i} => Ok({final_size}), // {op_name}")?;
} else {
let mut fixed_offset: usize = 2;
for f in &op.fields {
if f.is_array || f.name == "m_type" || f.name == "m_strict" {
continue;
}
let info = field_type_info(&f.ty);
fixed_offset = round_up(fixed_offset, info.align);
fixed_offset += info.size;
}
let minimum_length = round_up(fixed_offset, STRUCT_ALIGN);
let m_length_offset = find_m_length_offset(&op.fields);
let op_name = &op.name;
writeln!(w, " {i} => {{ // {op_name} (variable-length)")?;
writeln!(w, " let m_length_end = at + {m_length_offset} + 4;")?;
writeln!(w, " if m_length_end > bytes.len() {{")?;
writeln!(
w,
" return Err(ValidationErrorKind::TruncatedInstruction);"
)?;
writeln!(w, " }}")?;
writeln!(
w,
" let raw = u32::from_ne_bytes(bytes[at + {m_length_offset}..m_length_end].try_into().unwrap());"
)?;
writeln!(w, " if raw < {minimum_length} {{")?;
writeln!(w, " return Err(ValidationErrorKind::InvalidLength);")?;
writeln!(w, " }}")?;
writeln!(w, " Ok(raw as usize)")?;
writeln!(w, " }}")?;
}
}
writeln!(w, " _ => Err(ValidationErrorKind::UnknownOpcode),")?;
writeln!(w, " }}")?;
writeln!(w, "}}")?;
writeln!(w)?;
Ok(())
}
fn emit_scalar_field_check(
mut w: impl Write,
field_name: &str,
ty: &str,
offset: usize,
) -> Result<(), Box<dyn std::error::Error>> {
match ty {
"Operand" => writeln!(w, " validate_operand(read_u32(bytes, at + {offset}), ctx)?;")?,
"Optional<Operand>" => writeln!(
w,
" validate_optional_operand(read_u32(bytes, at + {offset}), ctx)?;"
)?,
"Label" => writeln!(w, " validate_label(read_u32(bytes, at + {offset}), ctx)?;")?,
"Optional<Label>" => {
// Encoded as 8 bytes: u32 value + u8 has_value + 3 bytes pad.
writeln!(w, " if bytes[at + {offset} + 4] != 0 {{")?;
writeln!(
w,
" validate_label(read_u32(bytes, at + {offset}), ctx)?;"
)?;
writeln!(w, " }}")?;
}
"IdentifierTableIndex" => writeln!(
w,
" validate_identifier_index(read_u32(bytes, at + {offset}), ctx)?;"
)?,
"Optional<IdentifierTableIndex>" => writeln!(
w,
" validate_optional_identifier_index(read_u32(bytes, at + {offset}), ctx)?;"
)?,
"StringTableIndex" => writeln!(
w,
" validate_string_index(read_u32(bytes, at + {offset}), ctx)?;"
)?,
"Optional<StringTableIndex>" => writeln!(
w,
" validate_optional_string_index(read_u32(bytes, at + {offset}), ctx)?;"
)?,
"PropertyKeyTableIndex" => writeln!(
w,
" validate_property_key_index(read_u32(bytes, at + {offset}), ctx)?;"
)?,
"RegexTableIndex" => {
// The regex table is not consulted at runtime; skip range-checking.
}
"PropertyLookupCacheIndex" => writeln!(
w,
" validate_property_lookup_cache_index(read_u32(bytes, at + {offset}), ctx)?;"
)?,
"GlobalVariableCacheIndex" => writeln!(
w,
" validate_global_variable_cache_index(read_u32(bytes, at + {offset}), ctx)?;"
)?,
"EnvironmentCoordinateCacheIndex" => writeln!(
w,
" validate_environment_coordinate_cache_index(read_u32(bytes, at + {offset}), ctx)?;"
)?,
"TemplateObjectCacheIndex" => writeln!(
w,
" validate_template_object_cache_index(read_u32(bytes, at + {offset}), ctx)?;"
)?,
"ObjectShapeCacheIndex" => writeln!(
w,
" validate_object_shape_cache_index(read_u32(bytes, at + {offset}), ctx)?;"
)?,
"ObjectPropertyIteratorCacheIndex" => writeln!(
w,
" validate_object_property_iterator_cache_index(read_u32(bytes, at + {offset}), ctx)?;"
)?,
"u32" => {
// The .def gives us no first-class types for SFD, class-blueprint,
// or object-shape cache references stored as u32. Recognize the
// canonical field names so these still get range-checked.
if field_name == "m_shared_function_data_index" {
writeln!(
w,
" validate_shared_function_data_index(read_u32(bytes, at + {offset}), ctx)?;"
)?;
} else if field_name == "m_class_blueprint_index" {
writeln!(
w,
" validate_class_blueprint_index(read_u32(bytes, at + {offset}), ctx)?;"
)?;
} else if field_name == "m_shape_cache_index" {
writeln!(
w,
" validate_object_shape_cache_index(read_u32(bytes, at + {offset}), ctx)?;"
)?;
}
}
"Completion::Type" => writeln!(
w,
" validate_completion_type(read_u32(bytes, at + {offset}), ctx)?;"
)?,
"IteratorHint" => writeln!(
w,
" validate_iterator_hint(read_u32(bytes, at + {offset}), ctx)?;"
)?,
"EnvironmentMode" => writeln!(
w,
" validate_environment_mode(read_u32(bytes, at + {offset}), ctx)?;"
)?,
"PutKind" => writeln!(
w,
" validate_put_kind(read_u32(bytes, at + {offset}), ctx)?;"
)?,
"ArgumentsKind" => writeln!(
w,
" validate_arguments_kind(read_u32(bytes, at + {offset}), ctx)?;"
)?,
// bool, u64, Value, EnvironmentCoordinate, Builtin: no per-field
// bound applied here.
_ => {}
}
Ok(())
}
fn emit_array_elem_check(mut w: impl Write, ty: &str) -> Result<(), Box<dyn std::error::Error>> {
match ty {
"Operand" => writeln!(w, " validate_operand(read_u32(bytes, __off), ctx)?;")?,
"Optional<Operand>" => writeln!(
w,
" validate_optional_operand(read_u32(bytes, __off), ctx)?;"
)?,
"Value" => {
// Trailing Value array (NewPrimitiveArray): no per-element check;
// the count was already bounded against the instruction length.
}
other => panic!("Array element type not supported: {other}"),
}
Ok(())
}
fn generate_validate_instruction(mut w: impl Write, ops: &[OpDef]) -> Result<(), Box<dyn std::error::Error>> {
let layouts = compute_layouts(ops);
writeln!(
w,
"/// Per-opcode field validation, dispatched by Pass 2 of the validator."
)?;
writeln!(
w,
"pub fn validate_instruction(opcode: u8, ctx: &super::validator::ValidationContext, at: usize) -> Result<(), super::validator::ValidationErrorKind> {{"
)?;
writeln!(w, " use super::validator::*;")?;
writeln!(w, " let bytes = ctx.bytes;")?;
writeln!(w, " match opcode {{")?;
for (i, op) in ops.iter().enumerate() {
let layout = layouts.get(&op.name).expect("layout missing for op");
let arrays: Vec<&Field> = op.fields.iter().filter(|f| f.is_array).collect();
let has_array = !arrays.is_empty();
// Map each count field's name to the array it sizes, so we can skip
// the count field when emitting scalar checks (it gets read alongside
// the array bound below) and avoid duplicate work.
let mut count_field_names: std::collections::HashSet<String> = std::collections::HashSet::new();
for af in &arrays {
count_field_names.insert(find_count_field_name(op, af));
}
let op_name = &op.name;
writeln!(w, " {i} => {{ // {op_name}")?;
for f in &op.fields {
if f.is_array {
continue;
}
if f.name == "m_type" || f.name == "m_strict" || f.name == "m_length" {
continue;
}
if count_field_names.contains(&f.name) {
continue;
}
let offset = *layout.field_offsets.get(&f.name).expect("missing field offset");
emit_scalar_field_check(&mut w, &f.name, &f.ty, offset)?;
}
if has_array {
let m_length_offset = *layout
.field_offsets
.get("m_length")
.expect("variable-length op missing m_length");
// All variable-length ops in Bytecode.def carry exactly one trailing
// array; if that ever changes, this loop validates each independently.
for af in &arrays {
let array_offset = *layout.field_offsets.get(&af.name).expect("missing array offset");
let count_field = find_count_field_name(op, af);
let count_offset = *layout
.field_offsets
.get(&count_field)
.expect("missing count field offset");
let elem_size = field_type_info(&af.ty).size;
writeln!(
w,
" let __m_length = read_u32(bytes, at + {m_length_offset}) as usize;"
)?;
writeln!(
w,
" let __count = read_u32(bytes, at + {count_offset}) as usize;"
)?;
writeln!(
w,
" let __array_bytes = __count.saturating_mul({elem_size});"
)?;
writeln!(
w,
" if {array_offset}usize.saturating_add(__array_bytes) > __m_length {{"
)?;
writeln!(w, " return Err(ValidationErrorKind::InvalidLength);")?;
writeln!(w, " }}")?;
writeln!(w, " let __array_off = at + {array_offset};")?;
writeln!(w, " for __i in 0..__count {{")?;
writeln!(w, " let __off = __array_off + __i * {elem_size};")?;
emit_array_elem_check(&mut w, &af.ty)?;
writeln!(w, " }}")?;
}
}
writeln!(w, " }}")?;
}
writeln!(w, " _ => {{}}")?;
writeln!(w, " }}")?;
writeln!(w, " Ok(())")?;
writeln!(w, "}}")?;
writeln!(w)?;
Ok(())
}
fn generate_opcode_enum(mut w: impl Write, ops: &[OpDef]) -> Result<(), Box<dyn std::error::Error>> {
writeln!(
w,
"/// Bytecode opcode (u8), matching the C++ `Instruction::Type` enum."
)?;
writeln!(w, "#[derive(Debug, Clone, Copy, PartialEq, Eq)]")?;
writeln!(w, "#[repr(u8)]")?;
writeln!(w, "pub enum OpCode {{")?;
for (i, op) in ops.iter().enumerate() {
writeln!(w, " {} = {},", op.name, i)?;
}
writeln!(w, "}}")?;
writeln!(w)?;
Ok(())
}
fn generate_instruction_enum(mut w: impl Write, ops: &[OpDef]) -> Result<(), Box<dyn std::error::Error>> {
writeln!(w, "/// A bytecode instruction with typed fields.")?;
writeln!(w, "///")?;
writeln!(w, "/// Each variant corresponds to one C++ instruction class.")?;
writeln!(
w,
"/// During codegen, instructions are stored as these typed variants."
)?;
writeln!(
w,
"/// During flattening, they are serialized to bytes matching C++ layw."
)?;
writeln!(w, "#[derive(Debug, Clone)]")?;
writeln!(w, "pub enum Instruction {{")?;
for op in ops {
let fields = user_fields(op);
if fields.is_empty() {
writeln!(w, " {},", op.name)?;
} else {
writeln!(w, " {} {{", op.name)?;
for f in &fields {
let info = field_type_info(&f.ty);
let r_name = rust_field_name(&f.name);
if f.is_array {
writeln!(w, " {}: Vec<{}>,", r_name, info.rust_type)?;
} else {
writeln!(w, " {}: {},", r_name, info.rust_type)?;
}
}
writeln!(w, " }},")?;
}
}
writeln!(w, "}}")?;
writeln!(w)?;
Ok(())
}
fn generate_instruction_impl(mut w: impl Write, ops: &[OpDef]) -> Result<(), Box<dyn std::error::Error>> {
writeln!(w, "impl Instruction {{").unwrap();
generate_opcode_method(&mut w, ops)?;
generate_is_terminator_method(&mut w, ops)?;
generate_encode_method(&mut w, ops)?;
generate_encoded_size_method(&mut w, ops)?;
generate_visit_operands_method(&mut w, ops)?;
generate_visit_labels_method(&mut w, ops)?;
writeln!(w, " }}")?;
Ok(())
}
fn generate_opcode_method(mut w: impl Write, ops: &[OpDef]) -> Result<(), Box<dyn std::error::Error>> {
writeln!(w, " pub fn opcode(&self) -> OpCode {{")?;
writeln!(w, " match self {{")?;
for op in ops {
let fields = user_fields(op);
if fields.is_empty() {
writeln!(w, " Instruction::{} => OpCode::{},", op.name, op.name)?;
} else {
writeln!(
w,
" Instruction::{} {{ .. }} => OpCode::{},",
op.name, op.name
)?;
}
}
writeln!(w, " }}")?;
writeln!(w, " }}")?;
writeln!(w)?;
Ok(())
}
fn generate_is_terminator_method(mut w: impl Write, ops: &[OpDef]) -> Result<(), Box<dyn std::error::Error>> {
writeln!(w, " pub fn is_terminator(&self) -> bool {{")?;
writeln!(w, " matches!(self, ")?;
let terminators: Vec<&OpDef> = ops.iter().filter(|op| op.is_terminator).collect();
for (i, op) in terminators.iter().enumerate() {
let sep = if i + 1 < terminators.len() { " |" } else { "" };
let fields = user_fields(op);
if fields.is_empty() {
writeln!(w, " Instruction::{}{}", op.name, sep)?;
} else {
writeln!(w, " Instruction::{} {{ .. }}{}", op.name, sep)?;
}
}
writeln!(w, " )")?;
writeln!(w, " }}")?;
writeln!(w)?;
Ok(())
}
fn generate_encoded_size_method(mut w: impl Write, ops: &[OpDef]) -> Result<(), Box<dyn std::error::Error>> {
writeln!(w, " /// Returns the encoded size of this instruction in bytes.")?;
writeln!(w, " pub fn encoded_size(&self) -> usize {{")?;
writeln!(w, " match self {{")?;
for op in ops {
let fields = user_fields(op);
let has_array = op.fields.iter().any(|f| f.is_array);
if !has_array {
// Fixed-length: compute size statically
let mut offset: usize = 2; // header
for f in &op.fields {
if f.is_array || f.name == "m_type" || f.name == "m_strict" {
continue;
}
let info = field_type_info(&f.ty);
offset = round_up(offset, info.align);
offset += info.size;
}
let final_size = round_up(offset, 8);
let pat = if fields.is_empty() {
format!("Instruction::{}", op.name)
} else {
format!("Instruction::{} {{ .. }}", op.name)
};
writeln!(w, " {pat} => {final_size},")?;
} else {
// Variable-length: depends on array size
// Compute fixed part size
let mut fixed_offset: usize = 2;
for f in &op.fields {
if f.is_array || f.name == "m_type" || f.name == "m_strict" {
continue;
}
let info = field_type_info(&f.ty);
fixed_offset = round_up(fixed_offset, info.align);
fixed_offset += info.size;
}
// Find the array field and its element size
let Some(array_field) = op.fields.iter().find(|f| f.is_array) else {
continue;
};
let info = field_type_info(&array_field.ty);
let arr_name = rust_field_name(&array_field.name);
// C++ computes m_length as:
// round_up(alignof(void*), sizeof(*this) + sizeof(elem) * count)
// sizeof(*this) = round_up(fixed_offset, STRUCT_ALIGN) due to alignas(void*).
let sizeof_this = round_up(fixed_offset, STRUCT_ALIGN);
// Bind only the array field
let bindings: Vec<String> = fields
.iter()
.map(|f| {
let rname = rust_field_name(&f.name);
if rname == arr_name {
rname
} else {
format!("{rname}: _")
}
})
.collect();
writeln!(
w,
" Instruction::{} {{ {} }} => {{",
op.name,
bindings.join(", ")
)?;
writeln!(
w,
" let base = {} + {}.len() * {};",
sizeof_this, arr_name, info.size
)?;
writeln!(w, " (base + 7) & !7 // round up to 8")?;
writeln!(w, " }}")?;
}
}
writeln!(w, " }}")?;
writeln!(w, " }}")?;
writeln!(w)?;
Ok(())
}
fn generate_encode_method(mut w: impl Write, ops: &[OpDef]) -> Result<(), Box<dyn std::error::Error>> {
writeln!(
w,
" /// Encode this instruction into bytes matching the C++ struct layout."
)?;
writeln!(w, " pub fn encode(&self, strict: bool, buf: &mut Vec<u8>) {{")?;
writeln!(w, " let start = buf.len();")?;
writeln!(w, " match self {{")?;
for op in ops {
let fields = user_fields(op);
let has_array = op.fields.iter().any(|f| f.is_array);
let has_m_length = op.fields.iter().any(|f| f.name == "m_length");
// Generate match arm with field bindings
if fields.is_empty() {
writeln!(w, " Instruction::{} => {{", op.name)?;
} else {
let bindings: Vec<String> = fields.iter().map(|f| rust_field_name(&f.name)).collect();
writeln!(
w,
" Instruction::{} {{ {} }} => {{",
op.name,
bindings.join(", ")
)?;
}
// Write header: opcode (u8) + strict (u8) = 2 bytes
writeln!(w, " buf.push(OpCode::{} as u8);", op.name)?;
writeln!(w, " buf.push(strict as u8);")?;
// Track offset for C++ struct layw.
// We iterate ALL fields (including m_type, m_strict, m_length) for
// accurate alignment but only emit writes for user fields.
let mut offset: usize = 2;
// Iterate all non-array fields in declaration order
for f in &op.fields {
if f.is_array {
continue;
}
// m_type and m_strict are already written as the header
if f.name == "m_type" || f.name == "m_strict" {
continue;
}
let info = field_type_info(&f.ty);
// Pad to alignment
let aligned_offset = round_up(offset, info.align);
let pad = aligned_offset - offset;
if pad > 0 {
writeln!(w, " buf.extend_from_slice(&[0u8; {pad}]);")?;
}
offset = aligned_offset;
if f.name == "m_length" {
// Write placeholder (patched at end for variable-length instructions)
writeln!(
w,
" buf.extend_from_slice(&[0u8; 4]); // m_length placeholder"
)?;
} else {
let rname = rust_field_name(&f.name);
emit_field_write(&mut w, &rname, info.kind, false)?;
}
offset += info.size;
}
// Write trailing array elements
if has_array {
// sizeof(*this) in C++ = round_up(fixed_offset, STRUCT_ALIGN)
let sizeof_this = round_up(offset, STRUCT_ALIGN);
for f in &op.fields {
if !f.is_array {
continue;
}
let info = field_type_info(&f.ty);
let rname = rust_field_name(&f.name);
// Pad before first element if needed
let aligned_offset = round_up(offset, info.align);
let pad = aligned_offset - offset;
if pad > 0 {
writeln!(w, " buf.extend_from_slice(&[0u8; {pad}]);")?;
}
writeln!(w, " for item in {rname} {{")?;
emit_field_write(&mut w, "item", info.kind, true)?;
writeln!(w, " }}")?;
// Compute target size matching C++:
// round_up(STRUCT_ALIGN, sizeof(*this) + count * elem_size)
writeln!(
w,
" let target = ({} + {}.len() * {} + 7) & !7;",
sizeof_this, rname, info.size
)?;
writeln!(
w,
" while (buf.len() - start) < target {{ buf.push(0); }}"
)?;
}
if has_m_length {
// Patch m_length: it's the first u32 field after the header
let m_length_offset = find_m_length_offset(&op.fields);
writeln!(w, " let total_len = (buf.len() - start) as u32;")?;
writeln!(
w,
" buf[start + {}..start + {}].copy_from_slice(&total_len.to_ne_bytes());",
m_length_offset,
m_length_offset + 4
)?;
}
} else {
// Fixed-length: pad statically
let final_size = round_up(offset, 8);
let tail_pad = final_size - offset;
if tail_pad > 0 {
writeln!(w, " buf.extend_from_slice(&[0u8; {tail_pad}]);")?;
}
}
writeln!(w, " }}")?;
}
writeln!(w, " }}")?;
writeln!(w, " }}")?;
writeln!(w)?;
Ok(())
}
/// Emit code to write a field value into `w`.
///
/// All bindings from pattern matching and loop iteration are references (`&T`).
/// Rust auto-derefs for method calls, but explicit `*` is needed for casts
/// and direct pushes of Copy types.
fn emit_field_write(
mut w: impl Write,
name: &str,
kind: &str,
is_loop_item: bool,
) -> Result<(), Box<dyn std::error::Error>> {
let prefix = " ".repeat(if is_loop_item { 20 } else { 16 });
match kind {
"bool" => writeln!(w, "{prefix}buf.push(*{name} as u8);")?,
"u8" => writeln!(w, "{prefix}buf.push(*{name});")?,
"u32" => writeln!(w, "{prefix}buf.extend_from_slice(&{name}.to_ne_bytes());")?,
"u64" => writeln!(w, "{prefix}buf.extend_from_slice(&{name}.to_ne_bytes());")?,
"operand" => writeln!(w, "{prefix}buf.extend_from_slice(&{name}.raw().to_ne_bytes());")?,
"optional_operand" => {
writeln!(w, "{prefix}match {name} {{")?;
writeln!(
w,
"{prefix} Some(op) => buf.extend_from_slice(&op.raw().to_ne_bytes()),"
)?;
writeln!(
w,
"{prefix} None => buf.extend_from_slice(&Operand::INVALID.to_ne_bytes()),"
)?;
writeln!(w, "{prefix}}}")?;
}
"label" => writeln!(w, "{prefix}buf.extend_from_slice(&{name}.0.to_ne_bytes());")?,
"optional_label" => {
// C++ Optional<Label> layw: u32 value, bool has_value, 3 bytes padding = 8 bytes total
writeln!(w, "{prefix}match {name} {{")?;
writeln!(w, "{prefix} Some(lbl) => {{")?;
writeln!(w, "{prefix} buf.extend_from_slice(&lbl.0.to_ne_bytes());")?;
writeln!(w, "{prefix} buf.push(1); buf.push(0); buf.push(0); buf.push(0);")?;
writeln!(w, "{prefix} }}")?;
writeln!(w, "{prefix} None => {{")?;
writeln!(w, "{prefix} buf.extend_from_slice(&0u32.to_ne_bytes());")?;
writeln!(w, "{prefix} buf.push(0); buf.push(0); buf.push(0); buf.push(0);")?;
writeln!(w, "{prefix} }}")?;
writeln!(w, "{prefix}}}")?;
}
"u32_newtype" => writeln!(w, "{prefix}buf.extend_from_slice(&{name}.0.to_ne_bytes());")?,
"optional_u32_newtype" => {
writeln!(w, "{prefix}match {name} {{")?;
writeln!(
w,
"{prefix} Some(idx) => buf.extend_from_slice(&idx.0.to_ne_bytes()),"
)?;
writeln!(
w,
"{prefix} None => buf.extend_from_slice(&0xFFFF_FFFFu32.to_ne_bytes()),"
)?;
writeln!(w, "{prefix}}}")?;
}
"env_coord" => {
writeln!(w, "{prefix}buf.extend_from_slice(&{name}.hops.to_ne_bytes());")?;
writeln!(w, "{prefix}buf.extend_from_slice(&{name}.index.to_ne_bytes());")?;
}
other => panic!("Unknown encoding kind: {other}"),
}
Ok(())
}
fn generate_visit_operands_method(mut w: impl Write, ops: &[OpDef]) -> Result<(), Box<dyn std::error::Error>> {
writeln!(w, " /// Visit all `Operand` fields (for operand rewriting).")?;
writeln!(
w,
" pub fn visit_operands(&mut self, visitor: &mut dyn FnMut(&mut Operand)) {{"
)?;
writeln!(w, " match self {{")?;
for op in ops {
let fields = user_fields(op);
let operand_fields: Vec<&&Field> = fields
.iter()
.filter(|f| f.ty == "Operand" || f.ty == "Optional<Operand>")
.collect();
if operand_fields.is_empty() {
let pat = if fields.is_empty() {
format!("Instruction::{}", op.name)
} else {
format!("Instruction::{} {{ .. }}", op.name)
};
writeln!(w, " {pat} => {{}}")?;
continue;
}
// Bind the operand fields
let bindings: Vec<String> = fields
.iter()
.map(|f| {
let rname = rust_field_name(&f.name);
if f.ty == "Operand" || f.ty == "Optional<Operand>" {
rname
} else {
format!("{rname}: _")
}
})
.collect();
writeln!(
w,
" Instruction::{} {{ {} }} => {{",
op.name,
bindings.join(", ")
)?;
for f in &operand_fields {
let rname = rust_field_name(&f.name);
if f.is_array {
if f.ty == "Optional<Operand>" {
writeln!(
w,
" for op in {rname}.iter_mut().flatten() {{ visitor(op); }}"
)?;
} else {
writeln!(w, " for item in {rname}.iter_mut() {{ visitor(item); }}")?;
}
} else if f.ty == "Optional<Operand>" {
writeln!(w, " if let Some(op) = {rname} {{ visitor(op); }}")?;
} else {
writeln!(w, " visitor({rname});")?;
}
}
writeln!(w, " }}")?;
}
writeln!(w, " }}")?;
writeln!(w, " }}")?;
writeln!(w)?;
Ok(())
}
fn generate_visit_labels_method(mut w: impl Write, ops: &[OpDef]) -> Result<(), Box<dyn std::error::Error>> {
writeln!(w, " /// Visit all `Label` fields (for label linking).")?;
writeln!(
w,
" pub fn visit_labels(&mut self, visitor: &mut dyn FnMut(&mut Label)) {{"
)?;
writeln!(w, " match self {{")?;
for op in ops {
let fields = user_fields(op);
let label_fields: Vec<&&Field> = fields
.iter()
.filter(|f| f.ty == "Label" || f.ty == "Optional<Label>")
.collect();
if label_fields.is_empty() {
let pat = if fields.is_empty() {
format!("Instruction::{}", op.name)
} else {
format!("Instruction::{} {{ .. }}", op.name)
};
writeln!(w, " {pat} => {{}}")?;
continue;
}
let bindings: Vec<String> = fields
.iter()
.map(|f| {
let rname = rust_field_name(&f.name);
if f.ty == "Label" || f.ty == "Optional<Label>" {
rname
} else {
format!("{rname}: _")
}
})
.collect();
writeln!(
w,
" Instruction::{} {{ {} }} => {{",
op.name,
bindings.join(", ")
)?;
for f in &label_fields {
let rname = rust_field_name(&f.name);
if f.is_array {
if f.ty == "Optional<Label>" {
writeln!(w, " for item in {rname}.iter_mut() {{")?;
writeln!(w, " if let Some(lbl) = item {{ visitor(lbl); }}")?;
writeln!(w, " }}")?;
} else {
writeln!(w, " for item in {rname}.iter_mut() {{ visitor(item); }}")?;
}
} else if f.ty == "Optional<Label>" {
writeln!(w, " if let Some(lbl) = {rname} {{ visitor(lbl); }}")?;
} else {
writeln!(w, " visitor({rname});")?;
}
}
writeln!(w, " }}")?;
}
writeln!(w, " }}")?;
writeln!(w, " }}")?;
Ok(())
}
fn main() -> Result<(), Box<dyn std::error::Error>> {
let manifest_dir = PathBuf::from(env::var("CARGO_MANIFEST_DIR")?);
let def_path = manifest_dir.join("../Bytecode/Bytecode.def");
println!("cargo:rerun-if-changed={}", def_path.display());
println!("cargo:rerun-if-changed=build.rs");
println!("cargo:rerun-if-changed=cbindgen.toml");
println!("cargo:rerun-if-env-changed=FFI_OUTPUT_DIR");
println!("cargo:rerun-if-changed=src");
let out_dir = PathBuf::from(env::var("OUT_DIR")?);
let ffi_out_dir = env::var("FFI_OUTPUT_DIR")
.map(PathBuf::from)
.unwrap_or_else(|_| out_dir.clone());
cbindgen::generate(manifest_dir).map_or_else(
|error| match error {
cbindgen::Error::ParseSyntaxError { .. } => {}
e => panic!("{e:?}"),
},
|bindings| {
let header_path = out_dir.join("RustFFI.h");
bindings.write_to_file(&header_path);
if ffi_out_dir != out_dir {
bindings.write_to_file(ffi_out_dir.join("RustFFI.h"));
}
},
);
let file = fs::OpenOptions::new()
.write(true)
.create(true)
.truncate(true) // empties contents of the file
.open(out_dir.join("instruction_generated.rs"))?;
let content = fs::read_to_string(&def_path).expect("Failed to read Bytecode.def");
let ops = bytecode_def::parse_bytecode_def(&content);
generate_rust_code(file, &ops)
}