Version v3.0.0-beta.8

When searching for a symbol by name, only look at
symbols that are defined within the object,
ignoring extern symbols (symbols without section).

Fixes #180
Fixes #181

.cargo/config.toml

@@ -1,5 +1,4 @@
-[target.x86_64-pc-windows-msvc]
-linker = "rust-lld"
-
-[target.aarch64-pc-windows-msvc]
-linker = "rust-lld"
+# statically link the C runtime so the executable does not depend on
+# that shared/dynamic library.
+[target.'cfg(all(target_env = "msvc", target_os = "windows"))']
+rustflags = ["-C", "target-feature=+crt-static"]

Cargo.toml

@@ -14,7 +14,7 @@ strip = "debuginfo"
 codegen-units = 1
 
 [workspace.package]
-version = "3.0.0-beta.6"
+version = "3.0.0-beta.8"
 authors = ["Luke Street <luke@street.dev>"]
 edition = "2024"
 license = "MIT OR Apache-2.0"

README.md

@@ -16,11 +16,12 @@ Features:
 
 Supports:
 
-- PowerPC 750CL (GameCube, Wii)
-- MIPS (N64, PS1, PS2, PSP)
-- x86 (COFF only at the moment)
 - ARM (GBA, DS, 3DS)
-- ARM64 (Switch, experimental)
+- ARM64 (Switch)
+- MIPS (N64, PS1, PS2, PSP)
+- PowerPC (GameCube, Wii)
+- SuperH (Saturn, Dreamcast)
+- x86 (COFF only)
 
 See [Usage](#usage) for more information.
 

config.schema.json

@@ -175,6 +175,10 @@
           "type": "boolean",
           "description": "If true, objdiff will run the build command with the context file as an argument to generate it.",
           "default": false
+        },
+        "preset_id": {
+          "type": "number",
+          "description": "The decomp.me preset ID to use for the scratch.\nCompiler and flags in the config will take precedence over the preset, but the preset is useful for organizational purposes."
         }
       },
       "required": [

deny.toml

@@ -103,6 +103,7 @@ allow = [
     "0BSD",
     "OFL-1.1",
     "Ubuntu-font-1.0",
+    "CDLA-Permissive-2.0",
 ]
 # The confidence threshold for detecting a license from license text.
 # The higher the value, the more closely the license text must be to the

objdiff-cli/Cargo.toml

@@ -28,7 +28,7 @@ supports-color = "3.0"
 time = { version = "0.3", features = ["formatting", "local-offset"] }
 tracing = "0.1"
 tracing-subscriber = { version = "0.3", features = ["env-filter"] }
-typed-path = "0.10"
+typed-path = "0.11"
 
 [target.'cfg(target_env = "musl")'.dependencies]
 mimalloc = "0.1"

objdiff-cli/src/cmd/report.rs

@@ -85,6 +85,9 @@ pub fn run(args: Args) -> Result<()> {
 fn generate(args: GenerateArgs) -> Result<()> {
     let mut diff_config = diff::DiffObjConfig {
         function_reloc_diffs: diff::FunctionRelocDiffs::None,
+        combine_data_sections: true,
+        combine_text_sections: true,
+        ppc_calculate_pool_relocations: false,
         ..Default::default()
     };
     apply_config_args(&mut diff_config, &args.config)?;
@@ -225,6 +228,7 @@ fn report_object(
                 demangled_name: None,
                 virtual_address: section.virtual_address,
             }),
+            address: None,
         });
 
         match section.kind {
@@ -272,6 +276,7 @@ fn report_object(
                     demangled_name: symbol.demangled_name.clone(),
                     virtual_address: symbol.virtual_address,
                 }),
+                address: symbol.address.checked_sub(section.address),
             });
             if match_percent == 100.0 {
                 measures.matched_functions += 1;
@@ -279,6 +284,16 @@ fn report_object(
             measures.total_functions += 1;
         }
     }
+    sections.sort_by(|a, b| a.name.cmp(&b.name));
+    let reverse_fn_order = object.metadata.reverse_fn_order.unwrap_or(false);
+    functions.sort_by(|a, b| {
+        if reverse_fn_order {
+            b.address.unwrap_or(0).cmp(&a.address.unwrap_or(0))
+        } else {
+            a.address.unwrap_or(u64::MAX).cmp(&b.address.unwrap_or(u64::MAX))
+        }
+        .then_with(|| a.size.cmp(&b.size))
+    });
     if metadata.complete.unwrap_or(false) {
         measures.complete_code = measures.total_code;
         measures.complete_data = measures.total_data;

objdiff-cli/src/views/function_diff.rs

@@ -450,11 +450,11 @@ impl UiView for FunctionDiffUi {
 
     fn reload(&mut self, state: &AppState) -> Result<()> {
         let left_sym =
-            state.left_obj.as_ref().and_then(|(o, _)| find_function(o, &self.symbol_name));
+            state.left_obj.as_ref().and_then(|(o, _)| o.symbol_by_name(&self.symbol_name));
         let right_sym =
-            state.right_obj.as_ref().and_then(|(o, _)| find_function(o, &self.symbol_name));
+            state.right_obj.as_ref().and_then(|(o, _)| o.symbol_by_name(&self.symbol_name));
         let prev_sym =
-            state.prev_obj.as_ref().and_then(|(o, _)| find_function(o, &self.symbol_name));
+            state.prev_obj.as_ref().and_then(|(o, _)| o.symbol_by_name(&self.symbol_name));
         self.num_rows = match (
             get_symbol(state.left_obj.as_ref(), left_sym),
             get_symbol(state.right_obj.as_ref(), right_sym),
@@ -650,12 +650,3 @@ fn get_symbol(
     let sym = sym?;
     Some((obj, sym, &diff.symbols[sym]))
 }
-
-fn find_function(obj: &Object, name: &str) -> Option<usize> {
-    for (symbol_idx, symbol) in obj.symbols.iter().enumerate() {
-        if symbol.name == name {
-            return Some(symbol_idx);
-        }
-    }
-    None
-}

objdiff-core/Cargo.toml

@@ -133,7 +133,7 @@ prost = { version = "0.13", default-features = false, features = ["prost-derive"
 regex = { version = "1.11", default-features = false, features = [], optional = true }
 serde = { version = "1.0", default-features = false, features = ["derive"], optional = true }
 similar = { version = "2.7", default-features = false, features = ["hashbrown"], optional = true, git = "https://github.com/encounter/similar.git", branch = "no_std" }
-typed-path = { version = "0.10", default-features = false, optional = true }
+typed-path = { version = "0.11", default-features = false, optional = true }
 
 # config
 globset = { version = "0.4", default-features = false, optional = true }
@@ -158,7 +158,7 @@ iced-x86 = { version = "1.21", default-features = false, features = ["decoder",
 msvc-demangler = { version = "0.11", optional = true }
 
 # arm
-unarm = { version = "1.7", optional = true }
+unarm = { version = "1.8", optional = true }
 arm-attr = { version = "0.2", optional = true }
 
 # arm64
@@ -200,4 +200,4 @@ syn = { version = "2.0", optional = true }
 [dev-dependencies]
 # Enable all features for tests
 objdiff-core = { path = ".", features = ["all"] }
-insta = "1.42"
+insta = "1.43"

objdiff-core/README.md

@@ -5,11 +5,12 @@ objdiff-core contains the core functionality of [objdiff](https://github.com/enc
 ## Crate feature flags
 
 - **`all`**: Enables all main features.
+- **`bindings`**: Enables serialization and deserialization of objdiff data structures.
 - **`config`**: Enables objdiff configuration file support.
 - **`dwarf`**: Enables extraction of line number information from DWARF debug sections.
-- **`mips`**: Enables the MIPS backend powered by [rabbitizer](https://github.com/Decompollaborate/rabbitizer). (Note: C library with Rust bindings)
-- **`ppc`**: Enables the PowerPC backend powered by [ppc750cl](https://github.com/encounter/ppc750cl).
-- **`x86`**: Enables the x86 backend powered by [iced-x86](https://crates.io/crates/iced-x86).
-- **`arm`**: Enables the ARM backend powered by [unarm](https://github.com/AetiasHax/unarm).
 - **`arm64`**: Enables the ARM64 backend powered by [yaxpeax-arm](https://github.com/iximeow/yaxpeax-arm).
-- **`bindings`**: Enables serialization and deserialization of objdiff data structures.
+- **`arm`**: Enables the ARM backend powered by [unarm](https://github.com/AetiasHax/unarm).
+- **`mips`**: Enables the MIPS backend powered by [rabbitizer](https://github.com/Decompollaborate/rabbitizer).
+- **`ppc`**: Enables the PowerPC backend powered by [ppc750cl](https://github.com/encounter/ppc750cl).
+- **`superh`**: Enables the SuperH backend powered by an included disassembler.
+- **`x86`**: Enables the x86 backend powered by [iced-x86](https://crates.io/crates/iced-x86).

objdiff-core/protos/report.proto

@@ -99,6 +99,8 @@ message ReportItem {
   float fuzzy_match_percent = 3;
   // Extra metadata for this item
   optional ReportItemMetadata metadata = 4;
+  // Address of the item (section-relative offset)
+  optional uint64 address = 5;
 }
 
 // Extra metadata for an item

objdiff-core/src/arch/arm.rs

@@ -15,7 +15,7 @@ use crate::{
     diff::{ArmArchVersion, ArmR9Usage, DiffObjConfig, display::InstructionPart},
     obj::{
         InstructionRef, Relocation, RelocationFlags, ResolvedInstructionRef, ResolvedRelocation,
-        ScannedInstruction, Section, SectionKind, Symbol, SymbolFlag, SymbolFlagSet, SymbolKind,
+        Section, SectionKind, Symbol, SymbolFlag, SymbolFlagSet, SymbolKind,
     },
 };
 
@@ -187,14 +187,14 @@ impl Arch for ArchArm {
         self.disasm_modes = Self::get_mapping_symbols(sections, symbols);
     }
 
-    fn scan_instructions(
+    fn scan_instructions_internal(
         &self,
         address: u64,
         code: &[u8],
         section_index: usize,
         _relocations: &[Relocation],
         diff_config: &DiffObjConfig,
-    ) -> Result<Vec<ScannedInstruction>> {
+    ) -> Result<Vec<InstructionRef>> {
         let start_addr = address as u32;
         let end_addr = start_addr + code.len() as u32;
 
@@ -219,13 +219,13 @@ impl Arch for ArchArm {
         let mut next_mapping = mappings_iter.next();
 
         let ins_count = code.len() / mode.instruction_size(start_addr);
-        let mut ops = Vec::<ScannedInstruction>::with_capacity(ins_count);
+        let mut ops = Vec::<InstructionRef>::with_capacity(ins_count);
 
         let parse_flags = self.parse_flags(diff_config);
 
         let mut address = start_addr;
         while address < end_addr {
-            while let Some(next) = next_mapping.take_if(|x| address >= x.address) {
+            while let Some(next) = next_mapping.filter(|x| address >= x.address) {
                 // Change mapping
                 mode = next.mapping;
                 next_mapping = mappings_iter.next();
@@ -235,12 +235,10 @@ impl Arch for ArchArm {
             let data = &code[(address - start_addr) as usize..];
             if data.len() < ins_size {
                 // Push the remainder as data
-                ops.push(ScannedInstruction {
-                    ins_ref: InstructionRef {
-                        address: address as u64,
-                        size: data.len() as u8,
-                        opcode: u16::MAX,
-                    },
+                ops.push(InstructionRef {
+                    address: address as u64,
+                    size: data.len() as u8,
+                    opcode: u16::MAX,
                     branch_dest: None,
                 });
                 break;
@@ -256,12 +254,10 @@ impl Arch for ArchArm {
                 }
                 _ => {
                     // Invalid instruction size
-                    ops.push(ScannedInstruction {
-                        ins_ref: InstructionRef {
-                            address: address as u64,
-                            size: ins_size as u8,
-                            opcode: u16::MAX,
-                        },
+                    ops.push(InstructionRef {
+                        address: address as u64,
+                        size: ins_size as u8,
+                        opcode: u16::MAX,
                         branch_dest: None,
                     });
                     address += ins_size as u32;
@@ -325,8 +321,10 @@ impl Arch for ArchArm {
                 unarm::ParseMode::Data => (u16::MAX, None),
             };
 
-            ops.push(ScannedInstruction {
-                ins_ref: InstructionRef { address: address as u64, size: ins_size as u8, opcode },
+            ops.push(InstructionRef {
+                address: address as u64,
+                size: ins_size as u8,
+                opcode,
                 branch_dest: branch_dest.map(|x| x as u64),
             });
             address += ins_size as u32;

objdiff-core/src/arch/arm64.rs

@@ -18,7 +18,6 @@ use crate::{
     diff::{DiffObjConfig, display::InstructionPart},
     obj::{
         InstructionRef, Relocation, RelocationFlags, ResolvedInstructionRef, ResolvedRelocation,
-        ScannedInstruction,
     },
 };
 
@@ -30,16 +29,16 @@ impl ArchArm64 {
 }
 
 impl Arch for ArchArm64 {
-    fn scan_instructions(
+    fn scan_instructions_internal(
         &self,
         address: u64,
         code: &[u8],
         _section_index: usize,
         _relocations: &[Relocation],
         _diff_config: &DiffObjConfig,
-    ) -> Result<Vec<ScannedInstruction>> {
+    ) -> Result<Vec<InstructionRef>> {
         let start_address = address;
-        let mut ops = Vec::<ScannedInstruction>::with_capacity(code.len() / 4);
+        let mut ops = Vec::<InstructionRef>::with_capacity(code.len() / 4);
 
         let mut reader = U8Reader::new(code);
         let decoder = InstDecoder::default();
@@ -58,8 +57,10 @@ impl Arch for ArchArm64 {
                     DecodeError::InvalidOpcode
                     | DecodeError::InvalidOperand
                     | DecodeError::IncompleteDecoder => {
-                        ops.push(ScannedInstruction {
-                            ins_ref: InstructionRef { address, size: 4, opcode: u16::MAX },
+                        ops.push(InstructionRef {
+                            address,
+                            size: 4,
+                            opcode: u16::MAX,
                             branch_dest: None,
                         });
                         continue;
@@ -68,9 +69,9 @@ impl Arch for ArchArm64 {
             }
 
             let opcode = opcode_to_u16(ins.opcode);
-            let ins_ref = InstructionRef { address, size: 4, opcode };
-            let branch_dest = branch_dest(ins_ref, &code[offset as usize..offset as usize + 4]);
-            ops.push(ScannedInstruction { ins_ref, branch_dest });
+            let branch_dest =
+                branch_dest(opcode, address, &code[offset as usize..offset as usize + 4]);
+            ops.push(InstructionRef { address, size: 4, opcode, branch_dest });
         }
 
         Ok(ops)
@@ -163,7 +164,7 @@ impl Arch for ArchArm64 {
     }
 }
 
-fn branch_dest(ins_ref: InstructionRef, code: &[u8]) -> Option<u64> {
+fn branch_dest(opcode: u16, address: u64, code: &[u8]) -> Option<u64> {
     const OPCODE_B: u16 = opcode_to_u16(Opcode::B);
     const OPCODE_BL: u16 = opcode_to_u16(Opcode::BL);
     const OPCODE_BCC: u16 = opcode_to_u16(Opcode::Bcc(0));
@@ -173,21 +174,21 @@ fn branch_dest(ins_ref: InstructionRef, code: &[u8]) -> Option<u64> {
     const OPCODE_TBNZ: u16 = opcode_to_u16(Opcode::TBNZ);
 
     let word = u32::from_le_bytes(code.try_into().ok()?);
-    match ins_ref.opcode {
+    match opcode {
         OPCODE_B | OPCODE_BL => {
             let offset = ((word & 0x03ff_ffff) << 2) as i32;
             let extended_offset = (offset << 4) >> 4;
-            ins_ref.address.checked_add_signed(extended_offset as i64)
+            address.checked_add_signed(extended_offset as i64)
         }
         OPCODE_BCC | OPCODE_CBZ | OPCODE_CBNZ => {
             let offset = (word as i32 & 0x00ff_ffe0) >> 3;
             let extended_offset = (offset << 11) >> 11;
-            ins_ref.address.checked_add_signed(extended_offset as i64)
+            address.checked_add_signed(extended_offset as i64)
         }
         OPCODE_TBZ | OPCODE_TBNZ => {
             let offset = (word as i32 & 0x0007_ffe0) >> 3;
             let extended_offset = (offset << 16) >> 16;
-            ins_ref.address.checked_add_signed(extended_offset as i64)
+            address.checked_add_signed(extended_offset as i64)
         }
         _ => None,
     }

objdiff-core/src/arch/mips.rs

@@ -3,7 +3,6 @@ use alloc::{
     string::{String, ToString},
     vec::Vec,
 };
-use core::ops::Range;
 
 use anyhow::{Result, bail};
 use object::{Endian as _, Object as _, ObjectSection as _, ObjectSymbol as _, elf};
@@ -19,7 +18,7 @@ use crate::{
     diff::{DiffObjConfig, MipsAbi, MipsInstrCategory, display::InstructionPart},
     obj::{
         InstructionArg, InstructionArgValue, InstructionRef, Relocation, RelocationFlags,
-        ResolvedInstructionRef, ResolvedRelocation, ScannedInstruction, SymbolFlag, SymbolFlagSet,
+        ResolvedInstructionRef, ResolvedRelocation, SymbolFlag, SymbolFlagSet,
     },
 };
 
@@ -189,16 +188,16 @@ impl ArchMips {
 }
 
 impl Arch for ArchMips {
-    fn scan_instructions(
+    fn scan_instructions_internal(
         &self,
         address: u64,
         code: &[u8],
         _section_index: usize,
         _relocations: &[Relocation],
         diff_config: &DiffObjConfig,
-    ) -> Result<Vec<ScannedInstruction>> {
+    ) -> Result<Vec<InstructionRef>> {
         let instruction_flags = self.instruction_flags(diff_config);
-        let mut ops = Vec::<ScannedInstruction>::with_capacity(code.len() / 4);
+        let mut ops = Vec::<InstructionRef>::with_capacity(code.len() / 4);
         let mut cur_addr = address as u32;
         for chunk in code.chunks_exact(4) {
             let code = self.endianness.read_u32_bytes(chunk.try_into()?);
@@ -206,10 +205,7 @@ impl Arch for ArchMips {
                 rabbitizer::Instruction::new(code, Vram::new(cur_addr), instruction_flags);
             let opcode = instruction.opcode() as u16;
             let branch_dest = instruction.get_branch_vram_generic().map(|v| v.inner() as u64);
-            ops.push(ScannedInstruction {
-                ins_ref: InstructionRef { address: cur_addr as u64, size: 4, opcode },
-                branch_dest,
-            });
+            ops.push(InstructionRef { address: cur_addr as u64, size: 4, opcode, branch_dest });
             cur_addr += 4;
         }
         Ok(ops)
@@ -225,16 +221,7 @@ impl Arch for ArchMips {
         let display_flags = self.instruction_display_flags(diff_config);
         let opcode = instruction.opcode();
         cb(InstructionPart::opcode(opcode.name(), opcode as u16))?;
-        let start_address = resolved.symbol.address;
-        let function_range = start_address..start_address + resolved.symbol.size;
-        push_args(
-            &instruction,
-            resolved.relocation,
-            function_range,
-            resolved.section_index,
-            &display_flags,
-            cb,
-        )?;
+        push_args(&instruction, resolved.relocation, &display_flags, cb)?;
         Ok(())
     }
 
@@ -338,8 +325,6 @@ impl Arch for ArchMips {
 fn push_args(
     instruction: &rabbitizer::Instruction,
     relocation: Option<ResolvedRelocation>,
-    function_range: Range<u64>,
-    section_index: usize,
     display_flags: &rabbitizer::InstructionDisplayFlags,
     mut arg_cb: impl FnMut(InstructionPart) -> Result<()>,
 ) -> Result<()> {
@@ -362,23 +347,7 @@ fn push_args(
             }
             ValuedOperand::core_label(..) | ValuedOperand::core_branch_target_label(..) => {
                 if let Some(resolved) = relocation {
-                    // If the relocation target is within the current function, we can
-                    // convert it into a relative branch target. Note that we check
-                    // target_address > start_address instead of >= so that recursive
-                    // tail calls are not considered branch targets.
-                    let target_address =
-                        resolved.symbol.address.checked_add_signed(resolved.relocation.addend);
-                    if resolved.symbol.section == Some(section_index)
-                        && target_address.is_some_and(|addr| {
-                            addr > function_range.start && addr < function_range.end
-                        })
-                    {
-                        // TODO move this logic up a level
-                        let target_address = target_address.unwrap();
-                        arg_cb(InstructionPart::branch_dest(target_address))?;
-                    } else {
-                        push_reloc(resolved.relocation, &mut arg_cb)?;
-                    }
+                    push_reloc(resolved.relocation, &mut arg_cb)?;
                 } else if let Some(branch_dest) = instruction
                     .get_branch_offset_generic()
                     .map(|o| (instruction.vram() + o).inner() as u64)

objdiff-core/src/arch/mod.rs

@@ -11,8 +11,8 @@ use crate::{
         display::{ContextItem, HoverItem, InstructionPart},
     },
     obj::{
-        InstructionArg, Object, ParsedInstruction, Relocation, RelocationFlags,
-        ResolvedInstructionRef, ScannedInstruction, Section, Symbol, SymbolFlagSet, SymbolKind,
+        InstructionArg, InstructionRef, Object, ParsedInstruction, Relocation, RelocationFlags,
+        ResolvedInstructionRef, ResolvedSymbol, Section, Symbol, SymbolFlagSet, SymbolKind,
     },
     util::ReallySigned,
 };
@@ -182,46 +182,108 @@ impl DataType {
     }
 }
 
-pub trait Arch: Send + Sync + Debug {
-    // Finishes arch-specific initialization that must be done after sections have been combined.
-    fn post_init(&mut self, _sections: &[Section], _symbols: &[Symbol]) {}
-
+impl dyn Arch {
     /// Generate a list of instructions references (offset, size, opcode) from the given code.
     ///
-    /// The opcode IDs are used to generate the initial diff. Implementations should do as little
-    /// parsing as possible here: just enough to identify the base instruction opcode, size, and
-    /// possible branch destination (for visual representation). As needed, instructions are parsed
-    /// via `process_instruction` to compare their arguments.
-    fn scan_instructions(
+    /// See [`scan_instructions_internal`] for more details.
+    pub fn scan_instructions(
         &self,
-        address: u64,
-        code: &[u8],
-        section_index: usize,
-        relocations: &[Relocation],
+        resolved: ResolvedSymbol,
         diff_config: &DiffObjConfig,
-    ) -> Result<Vec<ScannedInstruction>>;
+    ) -> Result<Vec<InstructionRef>> {
+        let mut result = self.scan_instructions_internal(
+            resolved.symbol.address,
+            resolved.data,
+            resolved.section_index,
+            &resolved.section.relocations,
+            diff_config,
+        )?;
+
+        let function_start = resolved.symbol.address;
+        let function_end = function_start + resolved.symbol.size;
+
+        // Remove any branch destinations that are outside the function range
+        for ins in result.iter_mut() {
+            if let Some(branch_dest) = ins.branch_dest {
+                if branch_dest < function_start || branch_dest >= function_end {
+                    ins.branch_dest = None;
+                }
+            }
+        }
+
+        // Resolve relocation targets within the same function to branch destinations
+        let mut ins_iter = result.iter_mut().peekable();
+        'outer: for reloc in resolved
+            .section
+            .relocations
+            .iter()
+            .skip_while(|r| r.address < function_start)
+            .take_while(|r| r.address < function_end)
+        {
+            let ins = loop {
+                let Some(ins) = ins_iter.peek_mut() else {
+                    break 'outer;
+                };
+                if reloc.address < ins.address {
+                    continue 'outer;
+                }
+                let ins = ins_iter.next().unwrap();
+                if reloc.address >= ins.address && reloc.address < ins.address + ins.size as u64 {
+                    break ins;
+                }
+            };
+            // Clear existing branch destination for instructions with relocations
+            ins.branch_dest = None;
+            let Some(target) = resolved.obj.symbols.get(reloc.target_symbol) else {
+                continue;
+            };
+            if target.section != Some(resolved.section_index) {
+                continue;
+            }
+            let Some(target_address) = target.address.checked_add_signed(reloc.addend) else {
+                continue;
+            };
+            // If the target address is within the function range, set it as a branch destination
+            if target_address >= function_start && target_address < function_end {
+                ins.branch_dest = Some(target_address);
+            }
+        }
+
+        Ok(result)
+    }
 
     /// Parse an instruction to gather its mnemonic and arguments for more detailed comparison.
     ///
     /// This is called only when we need to compare the arguments of an instruction.
-    fn process_instruction(
+    pub fn process_instruction(
         &self,
         resolved: ResolvedInstructionRef,
         diff_config: &DiffObjConfig,
     ) -> Result<ParsedInstruction> {
         let mut mnemonic = None;
         let mut args = Vec::with_capacity(8);
+        let mut relocation_emitted = false;
         self.display_instruction(resolved, diff_config, &mut |part| {
             match part {
                 InstructionPart::Opcode(m, _) => mnemonic = Some(Cow::Owned(m.into_owned())),
-                InstructionPart::Arg(arg) => args.push(arg.into_static()),
+                InstructionPart::Arg(arg) => {
+                    if arg == InstructionArg::Reloc {
+                        relocation_emitted = true;
+                        // If the relocation was resolved to a branch destination, emit that instead.
+                        if let Some(dest) = resolved.ins_ref.branch_dest {
+                            args.push(InstructionArg::BranchDest(dest));
+                            return Ok(());
+                        }
+                    }
+                    args.push(arg.into_static());
+                }
                 _ => {}
             }
             Ok(())
         })?;
         // If the instruction has a relocation, but we didn't format it in the display, add it to
         // the end of the arguments list.
-        if resolved.relocation.is_some() && !args.contains(&InstructionArg::Reloc) {
+        if resolved.relocation.is_some() && !relocation_emitted {
             args.push(InstructionArg::Reloc);
         }
         Ok(ParsedInstruction {
@@ -230,6 +292,26 @@ pub trait Arch: Send + Sync + Debug {
             args,
         })
     }
+}
+
+pub trait Arch: Send + Sync + Debug {
+    /// Finishes arch-specific initialization that must be done after sections have been combined.
+    fn post_init(&mut self, _sections: &[Section], _symbols: &[Symbol]) {}
+
+    /// Generate a list of instructions references (offset, size, opcode) from the given code.
+    ///
+    /// The opcode IDs are used to generate the initial diff. Implementations should do as little
+    /// parsing as possible here: just enough to identify the base instruction opcode, size, and
+    /// possible branch destination (for visual representation). As needed, instructions are parsed
+    /// via `process_instruction` to compare their arguments.
+    fn scan_instructions_internal(
+        &self,
+        address: u64,
+        code: &[u8],
+        section_index: usize,
+        relocations: &[Relocation],
+        diff_config: &DiffObjConfig,
+    ) -> Result<Vec<InstructionRef>>;
 
     /// Format an instruction for display.
     ///
@@ -332,14 +414,14 @@ impl ArchDummy {
 }
 
 impl Arch for ArchDummy {
-    fn scan_instructions(
+    fn scan_instructions_internal(
         &self,
         _address: u64,
         _code: &[u8],
         _section_index: usize,
         _relocations: &[Relocation],
         _diff_config: &DiffObjConfig,
-    ) -> Result<Vec<ScannedInstruction>> {
+    ) -> Result<Vec<InstructionRef>> {
         Ok(Vec::new())
     }
 

objdiff-core/src/arch/ppc.rs

@@ -19,7 +19,7 @@ use crate::{
     },
     obj::{
         InstructionRef, Object, Relocation, RelocationFlags, ResolvedInstructionRef,
-        ResolvedRelocation, ScannedInstruction, Symbol, SymbolFlag, SymbolFlagSet,
+        ResolvedRelocation, Symbol, SymbolFlag, SymbolFlagSet,
     },
 };
 
@@ -82,24 +82,22 @@ impl ArchPpc {
 }
 
 impl Arch for ArchPpc {
-    fn scan_instructions(
+    fn scan_instructions_internal(
         &self,
         address: u64,
         code: &[u8],
         _section_index: usize,
         _relocations: &[Relocation],
         _diff_config: &DiffObjConfig,
-    ) -> Result<Vec<ScannedInstruction>> {
+    ) -> Result<Vec<InstructionRef>> {
         ensure!(code.len() & 3 == 0, "Code length must be a multiple of 4");
         let ins_count = code.len() / 4;
-        let mut insts = Vec::<ScannedInstruction>::with_capacity(ins_count);
+        let mut insts = Vec::<InstructionRef>::with_capacity(ins_count);
         for (cur_addr, ins) in ppc750cl::InsIter::new(code, address as u32) {
-            insts.push(ScannedInstruction {
-                ins_ref: InstructionRef {
-                    address: cur_addr as u64,
-                    size: 4,
-                    opcode: u8::from(ins.op) as u16,
-                },
+            insts.push(InstructionRef {
+                address: cur_addr as u64,
+                size: 4,
+                opcode: u8::from(ins.op) as u16,
                 branch_dest: ins.branch_dest(cur_addr).map(u64::from),
             });
         }

objdiff-core/src/arch/superh/disasm.rs

@@ -1,3 +1,5 @@
+use alloc::{format, vec::Vec};
+
 use crate::{diff::display::InstructionPart, obj::ResolvedInstructionRef};
 
 static REG_NAMES: [&str; 16] = [

objdiff-core/src/arch/superh/mod.rs

@@ -1,5 +1,4 @@
-use alloc::{string::String, vec::Vec};
-use std::collections::HashMap;
+use alloc::{collections::BTreeMap, format, string::String, vec, vec::Vec};
 
 use anyhow::{Result, bail};
 use object::elf;
@@ -7,9 +6,7 @@ use object::elf;
 use crate::{
     arch::{Arch, superh::disasm::sh2_disasm},
     diff::{DiffObjConfig, display::InstructionPart},
-    obj::{
-        InstructionRef, Relocation, RelocationFlags, ResolvedInstructionRef, ScannedInstruction,
-    },
+    obj::{InstructionRef, Relocation, RelocationFlags, ResolvedInstructionRef},
 };
 
 pub mod disasm;
@@ -27,15 +24,15 @@ struct DataInfo {
 }
 
 impl Arch for ArchSuperH {
-    fn scan_instructions(
+    fn scan_instructions_internal(
         &self,
         address: u64,
         code: &[u8],
         _section_index: usize,
         _relocations: &[Relocation],
         _diff_config: &DiffObjConfig,
-    ) -> Result<Vec<ScannedInstruction>> {
-        let mut ops = Vec::<ScannedInstruction>::with_capacity(code.len() / 2);
+    ) -> Result<Vec<InstructionRef>> {
+        let mut ops = Vec::<InstructionRef>::with_capacity(code.len() / 2);
         let mut offset = address;
 
         for chunk in code.chunks_exact(2) {
@@ -56,9 +53,7 @@ impl Arch for ArchSuperH {
                 Some(InstructionPart::Opcode(_, val)) => *val,
                 _ => 0,
             };
-            let ins_ref: InstructionRef =
-                InstructionRef { address: offset, size: 2, opcode: opcode_enum };
-            ops.push(ScannedInstruction { ins_ref, branch_dest });
+            ops.push(InstructionRef { address: offset, size: 2, opcode: opcode_enum, branch_dest });
             offset += 2;
         }
 
@@ -77,10 +72,12 @@ impl Arch for ArchSuperH {
 
         sh2_disasm(0, opcode, true, &mut parts, &resolved, &mut branch_dest);
 
-        if let Some(symbol_data) = resolved.section.symbol_data(resolved.symbol) {
+        if let Some(symbol_data) =
+            resolved.section.data_range(resolved.symbol.address, resolved.symbol.size as usize)
+        {
             // scan for data
             // map of instruction offsets to data target offsets
-            let mut data_offsets: HashMap<u64, DataInfo> = HashMap::<u64, DataInfo>::new();
+            let mut data_offsets = BTreeMap::<u64, DataInfo>::new();
 
             let mut pos: u64 = 0;
             for chunk in symbol_data.chunks_exact(2) {
@@ -255,7 +252,7 @@ mod test {
 
             arch.display_instruction(
                 ResolvedInstructionRef {
-                    ins_ref: InstructionRef { address: 0x1000, size: 2, opcode },
+                    ins_ref: InstructionRef { address: 0x1000, size: 2, opcode, branch_dest: None },
                     code: &code,
                     ..Default::default()
                 },
@@ -333,7 +330,7 @@ mod test {
 
             arch.display_instruction(
                 ResolvedInstructionRef {
-                    ins_ref: InstructionRef { address: 0x1000, size: 2, opcode },
+                    ins_ref: InstructionRef { address: 0x1000, size: 2, opcode, branch_dest: None },
                     code: &code,
                     ..Default::default()
                 },
@@ -416,7 +413,7 @@ mod test {
 
             arch.display_instruction(
                 ResolvedInstructionRef {
-                    ins_ref: InstructionRef { address: 0x1000, size: 2, opcode },
+                    ins_ref: InstructionRef { address: 0x1000, size: 2, opcode, branch_dest: None },
                     code: &code,
                     ..Default::default()
                 },
@@ -453,7 +450,7 @@ mod test {
 
             arch.display_instruction(
                 ResolvedInstructionRef {
-                    ins_ref: InstructionRef { address: 0x1000, size: 2, opcode },
+                    ins_ref: InstructionRef { address: 0x1000, size: 2, opcode, branch_dest: None },
                     code: &code,
                     ..Default::default()
                 },
@@ -502,7 +499,12 @@ mod test {
 
             arch.display_instruction(
                 ResolvedInstructionRef {
-                    ins_ref: InstructionRef { address: addr as u64, size: 2, opcode },
+                    ins_ref: InstructionRef {
+                        address: addr as u64,
+                        size: 2,
+                        opcode,
+                        branch_dest: None,
+                    },
                     code: &code,
                     ..Default::default()
                 },
@@ -538,7 +540,12 @@ mod test {
 
             arch.display_instruction(
                 ResolvedInstructionRef {
-                    ins_ref: InstructionRef { address: addr as u64, size: 2, opcode },
+                    ins_ref: InstructionRef {
+                        address: addr as u64,
+                        size: 2,
+                        opcode,
+                        branch_dest: None,
+                    },
                     code: &code,
                     ..Default::default()
                 },
@@ -577,7 +584,12 @@ mod test {
 
             arch.display_instruction(
                 ResolvedInstructionRef {
-                    ins_ref: InstructionRef { address: addr as u64, size: 2, opcode },
+                    ins_ref: InstructionRef {
+                        address: addr as u64,
+                        size: 2,
+                        opcode,
+                        branch_dest: None,
+                    },
                     code: &code,
                     ..Default::default()
                 },
@@ -616,7 +628,12 @@ mod test {
 
             arch.display_instruction(
                 ResolvedInstructionRef {
-                    ins_ref: InstructionRef { address: addr as u64, size: 2, opcode },
+                    ins_ref: InstructionRef {
+                        address: addr as u64,
+                        size: 2,
+                        opcode,
+                        branch_dest: None,
+                    },
                     code: &code,
                     ..Default::default()
                 },
@@ -648,7 +665,12 @@ mod test {
 
             arch.display_instruction(
                 ResolvedInstructionRef {
-                    ins_ref: InstructionRef { address: addr as u64, size: 2, opcode },
+                    ins_ref: InstructionRef {
+                        address: addr as u64,
+                        size: 2,
+                        opcode,
+                        branch_dest: None,
+                    },
                     code: &code,
                     ..Default::default()
                 },
@@ -677,7 +699,12 @@ mod test {
 
             arch.display_instruction(
                 ResolvedInstructionRef {
-                    ins_ref: InstructionRef { address: addr as u64, size: 2, opcode },
+                    ins_ref: InstructionRef {
+                        address: addr as u64,
+                        size: 2,
+                        opcode,
+                        branch_dest: None,
+                    },
                     code: &code,
                     ..Default::default()
                 },
@@ -709,7 +736,12 @@ mod test {
 
             arch.display_instruction(
                 ResolvedInstructionRef {
-                    ins_ref: InstructionRef { address: addr as u64, size: 2, opcode },
+                    ins_ref: InstructionRef {
+                        address: addr as u64,
+                        size: 2,
+                        opcode,
+                        branch_dest: None,
+                    },
                     code: &opcode.to_be_bytes(),
                     symbol: &Symbol {
                         address: 0x0606F378, // func base address
@@ -754,7 +786,12 @@ mod test {
 
             arch.display_instruction(
                 ResolvedInstructionRef {
-                    ins_ref: InstructionRef { address: addr as u64, size: 2, opcode },
+                    ins_ref: InstructionRef {
+                        address: addr as u64,
+                        size: 2,
+                        opcode,
+                        branch_dest: None,
+                    },
                     code: &opcode.to_be_bytes(),
                     symbol: &Symbol {
                         address: 0x0606F378, // func base address