fix(fuse): add --reproducible for byte-stable output (#325)

meld-cli/src/main.rs

@@ -76,6 +76,13 @@ enum Commands {
         #[arg(long)]
         no_attestation: bool,
 
+        /// Emit a byte-reproducible artifact (#325): derive the attestation
+        /// id from the output content and take the timestamp from
+        /// `SOURCE_DATE_EPOCH` (default epoch 0) instead of a random UUID +
+        /// wall clock, so identical input yields an identical sha256.
+        #[arg(long)]
+        reproducible: bool,
+
         /// Disable the `component-provenance` custom section (#192).
         /// The section maps each fused-module function index back to
         /// its originating component + function index; downstream
@@ -163,6 +170,7 @@ fn main() -> Result<()> {
             address_rebase,
             stats,
             no_attestation,
+            reproducible,
             no_component_provenance,
             dwarf,
             preserve_names,
@@ -178,6 +186,7 @@ fn main() -> Result<()> {
                 address_rebase,
                 stats,
                 no_attestation,
+                reproducible,
                 no_component_provenance,
                 dwarf,
                 preserve_names,
@@ -235,6 +244,7 @@ fn fuse_command(
     address_rebase: bool,
     show_stats: bool,
     no_attestation: bool,
+    reproducible: bool,
     no_component_provenance: bool,
     dwarf: String,
     preserve_names: bool,
@@ -333,6 +343,7 @@ fn fuse_command(
     let config = FuserConfig {
         memory_strategy,
         attestation: !no_attestation,
+        reproducible,
         component_provenance: !no_component_provenance,
         address_rebasing: address_rebase,
         preserve_names,

meld-core/benches/fusion_benchmarks.rs

@@ -72,6 +72,7 @@ fn bench_config() -> FuserConfig {
     FuserConfig {
         memory_strategy: MemoryStrategy::MultiMemory,
         attestation: false,
+        reproducible: false,
         component_provenance: false,
         address_rebasing: false,
         preserve_names: false,

meld-core/src/attestation.rs

@@ -123,6 +123,7 @@ pub struct FusionAttestationBuilder {
     tool_version: String,
     tool_hash: Option<String>,
     memory_strategy: String,
+    reproducible: bool,
 }
 
 impl FusionAttestationBuilder {
@@ -134,9 +135,18 @@ impl FusionAttestationBuilder {
             tool_version: tool_version.into(),
             tool_hash: None,
             memory_strategy: "shared".to_string(),
+            reproducible: false,
         }
     }
 
+    /// Emit a byte-reproducible attestation (#325): derive the id from the
+    /// output content and take the timestamp from `SOURCE_DATE_EPOCH`
+    /// (default epoch 0) instead of a random UUID + wall clock.
+    pub fn reproducible(mut self, reproducible: bool) -> Self {
+        self.reproducible = reproducible;
+        self
+    }
+
     /// Set the tool hash for reproducibility
     pub fn tool_hash(mut self, hash: impl Into<String>) -> Self {
         self.tool_hash = Some(hash.into());
@@ -211,8 +221,16 @@ impl FusionAttestationBuilder {
     /// Build the attestation
     pub fn build(self, output_bytes: &[u8], stats: &FusionStats) -> FusionAttestation {
         let output_hash = compute_sha256(output_bytes);
-        let timestamp = chrono_timestamp();
-        let attestation_id = generate_uuid();
+        // #325: reproducible mode derives the id from the output content and
+        // the timestamp from SOURCE_DATE_EPOCH so the artifact is byte-stable.
+        let (timestamp, attestation_id) = if self.reproducible {
+            (
+                chrono_timestamp_from(source_date_epoch()),
+                generate_uuid_from(entropy_from_hex(&output_hash)),
+            )
+        } else {
+            (chrono_timestamp(), generate_uuid())
+        };
 
         let size_reduction = if stats.input_size > 0 {
             let diff = stats.input_size as i128 - stats.output_size as i128;
@@ -328,6 +346,35 @@ pub(crate) fn generate_uuid_from(entropy: u128) -> String {
     )
 }
 
+/// Seconds-since-epoch for a reproducible attestation timestamp (#325).
+///
+/// Honors the ecosystem-standard `SOURCE_DATE_EPOCH` environment variable
+/// (a decimal seconds-since-Unix-epoch string); when it is unset or
+/// unparseable, falls back to `0` (`1970-01-01T00:00:00Z`) so a reproducible
+/// build is byte-stable regardless of the wall clock. Only consulted on the
+/// `--reproducible` path; normal builds keep the real `chrono_timestamp()`.
+pub(crate) fn source_date_epoch() -> u64 {
+    std::env::var("SOURCE_DATE_EPOCH")
+        .ok()
+        .and_then(|s| s.trim().parse::<u64>().ok())
+        .unwrap_or(0)
+}
+
+/// Derive deterministic UUID entropy from a content hash (#325).
+///
+/// Takes the fused output's SHA-256 hex string and folds its first 32 hex
+/// digits into a `u128`, so the attestation id is a stable function of the
+/// artifact content rather than the wall clock. Short/odd input degrades
+/// gracefully to whatever parses (never panics).
+pub(crate) fn entropy_from_hex(hash_hex: &str) -> u128 {
+    let head: String = hash_hex
+        .chars()
+        .filter(|c| c.is_ascii_hexdigit())
+        .take(32)
+        .collect();
+    u128::from_str_radix(&head, 16).unwrap_or(0)
+}
+
 /// Get current timestamp in ISO 8601 format using the system clock.
 ///
 /// Thin wrapper over [`chrono_timestamp_from`] sourcing seconds-since-epoch

meld-core/src/lib.rs

@@ -84,6 +84,12 @@ pub struct FuserConfig {
     /// Whether to generate attestation data
     pub attestation: bool,
 
+    /// Whether to emit a byte-reproducible attestation (#325): derive the
+    /// attestation id from the output content and source the timestamp from
+    /// `SOURCE_DATE_EPOCH` (default epoch 0) instead of a random UUID + the
+    /// wall clock, so identical input yields an identical fused artifact.
+    pub reproducible: bool,
+
     /// Whether to emit the `component-provenance` custom section
     /// (issue #192). When enabled (the default), every defined
     /// function in the fused module gets a back-pointer entry
@@ -144,6 +150,7 @@ impl Default for FuserConfig {
         Self {
             memory_strategy: MemoryStrategy::Auto,
             attestation: true,
+            reproducible: false,
             component_provenance: true,
             address_rebasing: false,
             preserve_names: false,
@@ -1832,7 +1839,8 @@ impl Fuser {
         stats: &FusionStats,
     ) -> attestation::FusionAttestation {
         let mut builder = FusionAttestationBuilder::new("meld", env!("CARGO_PKG_VERSION"))
-            .memory_strategy(self.memory_strategy_label());
+            .memory_strategy(self.memory_strategy_label())
+            .reproducible(self.config.reproducible);
 
         for (index, component) in self.components.iter().enumerate() {
             let name = component
@@ -1863,8 +1871,20 @@ impl Fuser {
         };
 
         let output_hash = attestation::compute_sha256(output_bytes);
-        let timestamp = attestation::chrono_timestamp();
-        let attestation_id = attestation::generate_uuid();
+        // #325: in reproducible mode, derive the id from the output content and
+        // the timestamp from SOURCE_DATE_EPOCH so the fused artifact is
+        // byte-stable across runs; otherwise keep the wall-clock/random id.
+        let (timestamp, attestation_id) = if self.config.reproducible {
+            (
+                attestation::chrono_timestamp_from(attestation::source_date_epoch()),
+                attestation::generate_uuid_from(attestation::entropy_from_hex(&output_hash)),
+            )
+        } else {
+            (
+                attestation::chrono_timestamp(),
+                attestation::generate_uuid(),
+            )
+        };
 
         let mut inputs = Vec::new();
         for (index, component) in self.components.iter().enumerate() {
@@ -2822,6 +2842,93 @@ mod tests {
         }
     }
 
+    /// #325: with `reproducible: true`, the fused artifact is byte-identical
+    /// across runs — the attestation id is derived from the output content and
+    /// the timestamp from `SOURCE_DATE_EPOCH` (default epoch 0), so the
+    /// random-UUID + wall-clock non-determinism is removed. The control
+    /// (reproducible off) must differ, proving the flag is what fixes it.
+    ///
+    /// Scoped to the default build (the shipped configuration). Under the
+    /// optional `attestation` (wsc) feature the emitted section additionally
+    /// carries `tool_parameters` / `metadata` as `wsc_attestation` `HashMap`
+    /// fields, whose serialization order meld cannot control from this side —
+    /// full byte-reproducibility on that path needs an upstream fix (sorted /
+    /// BTreeMap serialization). The default `FusionAttestationBuilder` path has
+    /// no such maps and is fully reproducible.
+    #[cfg(not(feature = "attestation"))]
+    #[test]
+    fn test_reproducible_attestation_is_byte_stable() {
+        use wasm_encoder::{
+            CodeSection, Component, ExportKind, ExportSection, Function, FunctionSection,
+            Instruction, MemorySection, MemoryType, Module as EncoderModule, ModuleSection,
+            TypeSection, ValType,
+        };
+        fn build_minimal_component() -> Vec<u8> {
+            let mut types = TypeSection::new();
+            types.ty().function([], [ValType::I32]);
+            let mut functions = FunctionSection::new();
+            functions.function(0);
+            let mut memory = MemorySection::new();
+            memory.memory(MemoryType {
+                minimum: 1,
+                maximum: None,
+                memory64: false,
+                shared: false,
+                page_size_log2: None,
+            });
+            let mut exports = ExportSection::new();
+            exports.export("run", ExportKind::Func, 0);
+            exports.export("memory", ExportKind::Memory, 0);
+            let mut code = CodeSection::new();
+            let mut func = Function::new([]);
+            func.instruction(&Instruction::I32Const(42));
+            func.instruction(&Instruction::End);
+            code.function(&func);
+            let mut module = EncoderModule::new();
+            module
+                .section(&types)
+                .section(&functions)
+                .section(&memory)
+                .section(&exports)
+                .section(&code);
+            let mut component = Component::new();
+            component.section(&ModuleSection(&module));
+            component.finish()
+        }
+        let component_bytes = build_minimal_component();
+
+        let fuse_once = |reproducible: bool| -> Vec<u8> {
+            let config = FuserConfig {
+                attestation: true,
+                reproducible,
+                ..FuserConfig::default()
+            };
+            let mut fuser = Fuser::new(config);
+            fuser
+                .add_component(&component_bytes)
+                .expect("add_component failed");
+            fuser.fuse().expect("fuse failed")
+        };
+
+        // Reproducible: identical bytes across independent runs.
+        let a = fuse_once(true);
+        let b = fuse_once(true);
+        assert_eq!(
+            a, b,
+            "#325: reproducible fusion must be byte-identical across runs"
+        );
+
+        // Control: without reproducible, the attestation's random UUID makes
+        // the output differ — otherwise the flag would be a no-op.
+        let c = fuse_once(false);
+        let d = fuse_once(false);
+        assert_ne!(
+            c, d,
+            "#325: non-reproducible fusion is expected to differ (random attestation id); \
+             if this ever holds, the reproducible flag is testing nothing"
+        );
+    }
+
     /// SR-20 / SC-8: Fail-fast when a core module (not a component) is passed
     /// to `add_component()`.
     ///

meld-core/tests/adapter_dwarf_e2e.rs

@@ -76,6 +76,7 @@ fn fuse_remap() -> Vec<u8> {
         memory_strategy: MemoryStrategy::MultiMemory,
         dwarf_handling: DwarfHandling::Remap,
         attestation: false,
+        reproducible: false,
         ..Default::default()
     });
     fuser.add_component_named(&a, Some("a")).unwrap();

meld-core/tests/adapter_safety.rs

@@ -431,6 +431,7 @@ fn test_sr12_adapter_generation_for_string_param() {
     let config = FuserConfig {
         memory_strategy: MemoryStrategy::MultiMemory,
         attestation: false,
+        reproducible: false,
         component_provenance: false,
         address_rebasing: false,
         preserve_names: false,
@@ -819,6 +820,7 @@ fn test_sr13_cabi_realloc_targets_correct_memory() {
     let config = FuserConfig {
         memory_strategy: MemoryStrategy::MultiMemory,
         attestation: false,
+        reproducible: false,
         component_provenance: false,
         address_rebasing: false,
         preserve_names: false,
@@ -1235,6 +1237,7 @@ fn test_sr15_list_copy_length() {
     let config = FuserConfig {
         memory_strategy: MemoryStrategy::MultiMemory,
         attestation: false,
+        reproducible: false,
         component_provenance: false,
         address_rebasing: false,
         preserve_names: false,
@@ -1731,6 +1734,7 @@ fn test_sr16_inner_pointer_fixup_list_string() {
     let config = FuserConfig {
         memory_strategy: MemoryStrategy::MultiMemory,
         attestation: false,
+        reproducible: false,
         component_provenance: false,
         address_rebasing: false,
         preserve_names: false,
@@ -2031,6 +2035,7 @@ fn test_sr17_utf8_to_utf16_string_transcoding() {
     let config = FuserConfig {
         memory_strategy: MemoryStrategy::MultiMemory,
         attestation: false,
+        reproducible: false,
         component_provenance: false,
         address_rebasing: false,
         preserve_names: false,
@@ -2120,6 +2125,7 @@ fn test_sr17_utf8_to_utf16_supplementary_plane_transcoding() {
     let config = FuserConfig {
         memory_strategy: MemoryStrategy::MultiMemory,
         attestation: false,
+        reproducible: false,
         component_provenance: false,
         address_rebasing: false,
         preserve_names: false,
@@ -2645,6 +2651,7 @@ fn test_sr17_utf16_to_utf8_supplementary_plane_transcoding() {
     let config = FuserConfig {
         memory_strategy: MemoryStrategy::MultiMemory,
         attestation: false,
+        reproducible: false,
         component_provenance: false,
         address_rebasing: false,
         preserve_names: false,
@@ -2719,6 +2726,7 @@ fn test_sr17_utf16_to_utf8_lone_high_surrogate_replacement() {
     let config = FuserConfig {
         memory_strategy: MemoryStrategy::MultiMemory,
         attestation: false,
+        reproducible: false,
         component_provenance: false,
         address_rebasing: false,
         preserve_names: false,
@@ -2783,6 +2791,7 @@ fn test_sr17_utf16_to_utf8_midstring_lone_surrogate_replacement() {
     let config = FuserConfig {
         memory_strategy: MemoryStrategy::MultiMemory,
         attestation: false,
+        reproducible: false,
         component_provenance: false,
         address_rebasing: false,
         preserve_names: false,
@@ -2872,6 +2881,7 @@ fn test_sr17_utf16_to_utf8_malformed_surrogate_matrix() {
         let config = FuserConfig {
             memory_strategy: MemoryStrategy::MultiMemory,
             attestation: false,
+            reproducible: false,
             component_provenance: false,
             address_rebasing: false,
             preserve_names: false,
@@ -3007,6 +3017,7 @@ fn ls_p_20_test_sr17_utf8_to_utf16_malformed_matrix() {
         let config = FuserConfig {
             memory_strategy: MemoryStrategy::MultiMemory,
             attestation: false,
+            reproducible: false,
             component_provenance: false,
             address_rebasing: false,
             preserve_names: false,
@@ -3082,6 +3093,7 @@ fn test_sr17_latin1_to_utf16_transcoding() {
         let config = FuserConfig {
             memory_strategy: MemoryStrategy::MultiMemory,
             attestation: false,
+            reproducible: false,
             component_provenance: false,
             address_rebasing: false,
             preserve_names: false,
@@ -3565,6 +3577,7 @@ fn fuse_run_i32(caller: &[u8], callee: &[u8], label: &str) -> i32 {
     let config = FuserConfig {
         memory_strategy: MemoryStrategy::MultiMemory,
         attestation: false,
+        reproducible: false,
         component_provenance: false,
         address_rebasing: false,
         preserve_names: false,