Implement diffing relocations within data sections (#154)

* Data view: Show data bytes with differing relocations as a diff

* Data view: Show differing relocations on hover

* Symbol list view: Adjust symbol/section match %s when relocations differ

* Improve data reloc diffing logic

* Don't make reloc diffs cause bytes to show as red or green

* Properly detect byte size of each relocation

* Data view: Add context menu for copying relocation target symbols

* Also show already-matching relocations on hover/right click

* Change font color for nonmatching relocs on hover

.github/workflows/build.yaml

@@ -65,7 +65,7 @@ jobs:
     continue-on-error: ${{ matrix.checks == 'advisories' }}
     steps:
       - uses: actions/checkout@v4
-      - uses: EmbarkStudios/cargo-deny-action@v1
+      - uses: EmbarkStudios/cargo-deny-action@v2
         with:
           command: check ${{ matrix.checks }}
 

Cargo.toml

@@ -13,9 +13,9 @@ strip = "debuginfo"
 codegen-units = 1
 
 [workspace.package]
-version = "2.5.0"
+version = "2.7.0"
 authors = ["Luke Street <luke@street.dev>"]
 edition = "2021"
 license = "MIT OR Apache-2.0"
 repository = "https://github.com/encounter/objdiff"
-rust-version = "1.74"
+rust-version = "1.81"

deny.toml

@@ -73,7 +73,6 @@ ignore = [
     #{ id = "RUSTSEC-0000-0000", reason = "you can specify a reason the advisory is ignored" },
     #"a-crate-that-is-yanked@0.1.1", # you can also ignore yanked crate versions if you wish
     #{ crate = "a-crate-that-is-yanked@0.1.1", reason = "you can specify why you are ignoring the yanked crate" },
-    { id = "RUSTSEC-2024-0384", reason = "Unmaintained indirect dependency" },
 ]
 # If this is true, then cargo deny will use the git executable to fetch advisory database.
 # If this is false, then it uses a built-in git library.
@@ -240,7 +239,7 @@ allow-git = []
 
 [sources.allow-org]
 # github.com organizations to allow git sources for
-github = ["notify-rs"]
+github = []
 # gitlab.com organizations to allow git sources for
 gitlab = []
 # bitbucket.org organizations to allow git sources for

objdiff-cli/src/cmd/diff.rs

@@ -12,7 +12,7 @@ use std::{
     time::Duration,
 };
 
-use anyhow::{bail, Context, Result};
+use anyhow::{anyhow, bail, Context, Result};
 use argp::FromArgs;
 use crossterm::{
     event,
@@ -27,9 +27,11 @@ use objdiff_core::{
         watcher::{create_watcher, Watcher},
         BuildConfig,
     },
-    config::{build_globset, default_watch_patterns, ProjectConfig, ProjectObject},
+    config::{build_globset, ProjectConfig, ProjectObject},
     diff,
-    diff::ObjDiff,
+    diff::{
+        ConfigEnum, ConfigPropertyId, ConfigPropertyKind, DiffObjConfig, MappingConfig, ObjDiff,
+    },
     jobs::{
         objdiff::{start_build, ObjDiffConfig},
         Job, JobQueue, JobResult,
@@ -63,9 +65,6 @@ pub struct Args {
     #[argp(option, short = 'u')]
     /// Unit name within project
     unit: Option<String>,
-    #[argp(switch, short = 'x')]
-    /// Relax relocation diffs
-    relax_reloc_diffs: bool,
     #[argp(option, short = 'o')]
     /// Output file (one-shot mode) ("-" for stdout)
     output: Option<PathBuf>,
@@ -75,6 +74,18 @@ pub struct Args {
     #[argp(positional)]
     /// Function symbol to diff
     symbol: Option<String>,
+    #[argp(option, short = 'c')]
+    /// Configuration property (key=value)
+    config: Vec<String>,
+    #[argp(option, short = 'm')]
+    /// Symbol mapping (target=base)
+    mapping: Vec<String>,
+    #[argp(option)]
+    /// Left symbol name for selection
+    selecting_left: Option<String>,
+    #[argp(option)]
+    /// Right symbol name for selection
+    selecting_right: Option<String>,
 }
 
 pub fn run(args: Args) -> Result<()> {
@@ -84,7 +95,9 @@ pub fn run(args: Args) -> Result<()> {
         &args.project,
         &args.unit,
     ) {
-        (Some(t), Some(b), None, None) => (Some(t.clone()), Some(b.clone()), None),
+        (Some(_), Some(_), None, None)
+        | (Some(_), None, None, None)
+        | (None, Some(_), None, None) => (args.target.clone(), args.base.clone(), None),
         (None, None, p, u) => {
             let project = match p {
                 Some(project) => project.clone(),
@@ -193,6 +206,43 @@ pub fn run(args: Args) -> Result<()> {
     }
 }
 
+fn build_config_from_args(args: &Args) -> Result<(DiffObjConfig, MappingConfig)> {
+    let mut diff_config = DiffObjConfig::default();
+    for config in &args.config {
+        let (key, value) = config.split_once('=').context("--config expects \"key=value\"")?;
+        let property_id = ConfigPropertyId::from_str(key)
+            .map_err(|()| anyhow!("Invalid configuration property: {}", key))?;
+        diff_config.set_property_value_str(property_id, value).map_err(|()| {
+            let mut options = String::new();
+            match property_id.kind() {
+                ConfigPropertyKind::Boolean => {
+                    options = "true, false".to_string();
+                }
+                ConfigPropertyKind::Choice(variants) => {
+                    for (i, variant) in variants.iter().enumerate() {
+                        if i > 0 {
+                            options.push_str(", ");
+                        }
+                        options.push_str(variant.value);
+                    }
+                }
+            }
+            anyhow!("Invalid value for {}. Expected one of: {}", property_id.name(), options)
+        })?;
+    }
+    let mut mapping_config = MappingConfig {
+        mappings: Default::default(),
+        selecting_left: args.selecting_left.clone(),
+        selecting_right: args.selecting_right.clone(),
+    };
+    for mapping in &args.mapping {
+        let (target, base) =
+            mapping.split_once('=').context("--mapping expects \"target=base\"")?;
+        mapping_config.mappings.insert(target.to_string(), base.to_string());
+    }
+    Ok((diff_config, mapping_config))
+}
+
 fn run_oneshot(
     args: &Args,
     output: &Path,
@@ -200,17 +250,19 @@ fn run_oneshot(
     base_path: Option<&Path>,
 ) -> Result<()> {
     let output_format = OutputFormat::from_option(args.format.as_deref())?;
-    let config = diff::DiffObjConfig {
-        relax_reloc_diffs: args.relax_reloc_diffs,
-        ..Default::default() // TODO
-    };
+    let (diff_config, mapping_config) = build_config_from_args(args)?;
     let target = target_path
-        .map(|p| obj::read::read(p, &config).with_context(|| format!("Loading {}", p.display())))
+        .map(|p| {
+            obj::read::read(p, &diff_config).with_context(|| format!("Loading {}", p.display()))
+        })
         .transpose()?;
     let base = base_path
-        .map(|p| obj::read::read(p, &config).with_context(|| format!("Loading {}", p.display())))
+        .map(|p| {
+            obj::read::read(p, &diff_config).with_context(|| format!("Loading {}", p.display()))
+        })
         .transpose()?;
-    let result = diff::diff_objs(&config, target.as_ref(), base.as_ref(), None)?;
+    let result =
+        diff::diff_objs(&diff_config, &mapping_config, target.as_ref(), base.as_ref(), None)?;
     let left = target.as_ref().and_then(|o| result.left.as_ref().map(|d| (o, d)));
     let right = base.as_ref().and_then(|o| result.right.as_ref().map(|d| (o, d)));
     write_output(&DiffResult::new(left, right), Some(output), output_format)?;
@@ -229,9 +281,10 @@ pub struct AppState {
     pub prev_obj: Option<(ObjInfo, ObjDiff)>,
     pub reload_time: Option<time::OffsetDateTime>,
     pub time_format: Vec<time::format_description::FormatItem<'static>>,
-    pub relax_reloc_diffs: bool,
     pub watcher: Option<Watcher>,
     pub modified: Arc<AtomicBool>,
+    pub diff_obj_config: DiffObjConfig,
+    pub mapping_config: MappingConfig,
 }
 
 fn create_objdiff_config(state: &AppState) -> ObjDiffConfig {
@@ -257,13 +310,8 @@ fn create_objdiff_config(state: &AppState) -> ObjDiffConfig {
             .is_some_and(|p| p.build_target.unwrap_or(false)),
         target_path: state.target_path.clone(),
         base_path: state.base_path.clone(),
-        diff_obj_config: diff::DiffObjConfig {
-            relax_reloc_diffs: state.relax_reloc_diffs,
-            ..Default::default() // TODO
-        },
-        symbol_mappings: Default::default(),
-        selecting_left: None,
-        selecting_right: None,
+        diff_obj_config: state.diff_obj_config.clone(),
+        mapping_config: state.mapping_config.clone(),
     }
 }
 
@@ -314,6 +362,7 @@ fn run_interactive(
     let Some(symbol_name) = &args.symbol else { bail!("Interactive mode requires a symbol name") };
     let time_format = time::format_description::parse_borrowed::<2>("[hour]:[minute]:[second]")
         .context("Failed to parse time format")?;
+    let (diff_obj_config, mapping_config) = build_config_from_args(&args)?;
     let mut state = AppState {
         jobs: Default::default(),
         waker: Default::default(),
@@ -326,17 +375,13 @@ fn run_interactive(
         prev_obj: None,
         reload_time: None,
         time_format,
-        relax_reloc_diffs: args.relax_reloc_diffs,
         watcher: None,
         modified: Default::default(),
+        diff_obj_config,
+        mapping_config,
     };
-    if let Some(project_dir) = &state.project_dir {
-        let watch_patterns = state
-            .project_config
-            .as_ref()
-            .and_then(|c| c.watch_patterns.as_ref())
-            .cloned()
-            .unwrap_or_else(default_watch_patterns);
+    if let (Some(project_dir), Some(project_config)) = (&state.project_dir, &state.project_config) {
+        let watch_patterns = project_config.build_watch_patterns()?;
         state.watcher = Some(create_watcher(
             state.modified.clone(),
             project_dir,

objdiff-cli/src/cmd/report.rs

@@ -169,22 +169,29 @@ fn report_object(
         }
         _ => {}
     }
-    let config = diff::DiffObjConfig { relax_reloc_diffs: true, ..Default::default() };
+    let diff_config = diff::DiffObjConfig {
+        function_reloc_diffs: diff::FunctionRelocDiffs::None,
+        ..Default::default()
+    };
+    let mapping_config = diff::MappingConfig::default();
     let target = object
         .target_path
         .as_ref()
         .map(|p| {
-            obj::read::read(p, &config).with_context(|| format!("Failed to open {}", p.display()))
+            obj::read::read(p, &diff_config)
+                .with_context(|| format!("Failed to open {}", p.display()))
         })
         .transpose()?;
     let base = object
         .base_path
         .as_ref()
         .map(|p| {
-            obj::read::read(p, &config).with_context(|| format!("Failed to open {}", p.display()))
+            obj::read::read(p, &diff_config)
+                .with_context(|| format!("Failed to open {}", p.display()))
         })
         .transpose()?;
-    let result = diff::diff_objs(&config, target.as_ref(), base.as_ref(), None)?;
+    let result =
+        diff::diff_objs(&diff_config, &mapping_config, target.as_ref(), base.as_ref(), None)?;
 
     let metadata = ReportUnitMetadata {
         complete: object.complete(),

objdiff-cli/src/views/function_diff.rs

@@ -3,7 +3,7 @@ use crossterm::event::{Event, KeyCode, KeyEventKind, KeyModifiers, MouseButton,
 use objdiff_core::{
     diff::{
         display::{display_diff, DiffText, HighlightKind},
-        ObjDiff, ObjInsDiffKind, ObjSymbolDiff,
+        FunctionRelocDiffs, ObjDiff, ObjInsDiffKind, ObjSymbolDiff,
     },
     obj::{ObjInfo, ObjSectionKind, ObjSymbol, SymbolRef},
 };
@@ -368,9 +368,15 @@ impl UiView for FunctionDiffUi {
                         self.scroll_x = self.scroll_x.saturating_sub(1);
                         result.redraw = true;
                     }
-                    // Toggle relax relocation diffs
+                    // Cycle through function relocation diff mode
                     KeyCode::Char('x') => {
-                        state.relax_reloc_diffs = !state.relax_reloc_diffs;
+                        state.diff_obj_config.function_reloc_diffs =
+                            match state.diff_obj_config.function_reloc_diffs {
+                                FunctionRelocDiffs::None => FunctionRelocDiffs::NameAddress,
+                                FunctionRelocDiffs::NameAddress => FunctionRelocDiffs::DataValue,
+                                FunctionRelocDiffs::DataValue => FunctionRelocDiffs::All,
+                                FunctionRelocDiffs::All => FunctionRelocDiffs::None,
+                            };
                         result.redraw = true;
                         return EventControlFlow::Reload;
                     }

objdiff-core/Cargo.toml

@@ -16,18 +16,104 @@ documentation = "https://docs.rs/objdiff-core"
 crate-type = ["cdylib", "rlib"]
 
 [features]
-all = ["config", "dwarf", "mips", "ppc", "x86", "arm", "arm64", "bindings", "build"]
-any-arch = ["config", "dep:bimap", "dep:strum", "dep:similar", "dep:flagset", "dep:log", "dep:memmap2", "dep:byteorder", "dep:num-traits"] # Implicit, used to check if any arch is enabled
-bindings = ["dep:serde_json", "dep:prost", "dep:pbjson", "dep:serde", "dep:prost-build", "dep:pbjson-build"]
-build = ["dep:shell-escape", "dep:path-slash", "dep:winapi", "dep:notify", "dep:notify-debouncer-full", "dep:reqwest", "dep:self_update", "dep:tempfile", "dep:time"]
-config = ["dep:bimap", "dep:globset", "dep:semver", "dep:serde_json", "dep:serde_yaml", "dep:serde", "dep:filetime"]
+all = [
+    # Features
+    "bindings",
+    "build",
+    "config",
+    "dwarf",
+    # Architectures
+    "mips",
+    "ppc",
+    "x86",
+    "arm",
+    "arm64",
+]
+# Implicit, used to check if any arch is enabled
+any-arch = [
+    "config",
+    "dep:bimap",
+    "dep:byteorder",
+    "dep:flagset",
+    "dep:heck",
+    "dep:log",
+    "dep:memmap2",
+    "dep:num-traits",
+    "dep:prettyplease",
+    "dep:proc-macro2",
+    "dep:quote",
+    "dep:serde",
+    "dep:serde_json",
+    "dep:similar",
+    "dep:strum",
+    "dep:syn",
+]
+bindings = [
+    "dep:pbjson",
+    "dep:pbjson-build",
+    "dep:prost",
+    "dep:prost-build",
+    "dep:serde",
+    "dep:serde_json",
+]
+build = [
+    "dep:notify",
+    "dep:notify-debouncer-full",
+    "dep:path-slash",
+    "dep:reqwest",
+    "dep:self_update",
+    "dep:shell-escape",
+    "dep:tempfile",
+    "dep:time",
+    "dep:winapi",
+]
+config = [
+    "dep:bimap",
+    "dep:filetime",
+    "dep:globset",
+    "dep:semver",
+    "dep:serde",
+    "dep:serde_json",
+    "dep:serde_yaml",
+]
 dwarf = ["dep:gimli"]
-mips = ["any-arch", "dep:rabbitizer"]
-ppc = ["any-arch", "dep:cwdemangle", "dep:cwextab", "dep:ppc750cl"]
-x86 = ["any-arch", "dep:cpp_demangle", "dep:iced-x86", "dep:msvc-demangler"]
-arm = ["any-arch", "dep:cpp_demangle", "dep:unarm", "dep:arm-attr"]
-arm64 = ["any-arch", "dep:cpp_demangle", "dep:yaxpeax-arch", "dep:yaxpeax-arm"]
-wasm = ["bindings", "any-arch", "dep:console_error_panic_hook", "dep:console_log", "dep:wasm-bindgen", "dep:tsify-next", "dep:log"]
+mips = [
+    "any-arch",
+    "dep:rabbitizer",
+]
+ppc = [
+    "any-arch",
+    "dep:cwdemangle",
+    "dep:cwextab",
+    "dep:ppc750cl",
+]
+x86 = [
+    "any-arch",
+    "dep:cpp_demangle",
+    "dep:iced-x86",
+    "dep:msvc-demangler",
+]
+arm = [
+    "any-arch",
+    "dep:arm-attr",
+    "dep:cpp_demangle",
+    "dep:unarm",
+]
+arm64 = [
+    "any-arch",
+    "dep:cpp_demangle",
+    "dep:yaxpeax-arch",
+    "dep:yaxpeax-arm",
+]
+wasm = [
+    "any-arch",
+    "bindings",
+    "dep:console_error_panic_hook",
+    "dep:console_log",
+    "dep:log",
+    "dep:tsify-next",
+    "dep:wasm-bindgen",
+]
 
 [package.metadata.docs.rs]
 features = ["all"]
@@ -63,7 +149,7 @@ gimli = { version = "0.31", default-features = false, features = ["read-all"], o
 
 # ppc
 cwdemangle = { version = "1.0", optional = true }
-cwextab = { version = "1.0.2", optional = true }
+cwextab = { version = "1.0", optional = true }
 ppc750cl = { version = "0.3", optional = true }
 
 # mips
@@ -83,10 +169,10 @@ yaxpeax-arch = { version = "0.3", default-features = false, features = ["std"],
 yaxpeax-arm = { version = "0.3", default-features = false, features = ["std"], optional = true }
 
 # build
-notify = { git = "https://github.com/notify-rs/notify", rev = "128bf6230c03d39dbb7f301ff7b20e594e34c3a2", optional = true }
-notify-debouncer-full = { git = "https://github.com/notify-rs/notify", rev = "128bf6230c03d39dbb7f301ff7b20e594e34c3a2", optional = true }
+notify = { version = "8.0.0", optional = true }
+notify-debouncer-full = { version = "0.5.0", optional = true }
 shell-escape = { version = "0.1", optional = true }
-tempfile = { version = "3.14", optional = true }
+tempfile = { version = "3.15", optional = true }
 time = { version = "0.3", optional = true }
 
 [target.'cfg(windows)'.dependencies]
@@ -96,13 +182,20 @@ winapi = { version = "0.3", optional = true }
 # For Linux static binaries, use rustls
 [target.'cfg(target_os = "linux")'.dependencies]
 reqwest = { version = "0.12", default-features = false, features = ["blocking", "json", "multipart", "rustls-tls"], optional = true }
-self_update = { version = "0.41", default-features = false, features = ["rustls"], optional = true }
+self_update = { version = "0.42", default-features = false, features = ["rustls"], optional = true }
 
 # For all other platforms, use native TLS
 [target.'cfg(not(target_os = "linux"))'.dependencies]
 reqwest = { version = "0.12", default-features = false, features = ["blocking", "json", "multipart", "default-tls"], optional = true }
-self_update = { version = "0.41", optional = true }
+self_update = { version = "0.42", optional = true }
 
 [build-dependencies]
-prost-build = { version = "0.13", optional = true }
+heck = { version = "0.5", optional = true }
 pbjson-build = { version = "0.7", optional = true }
+prettyplease = { version = "0.2", optional = true }
+proc-macro2 = { version = "1.0", optional = true }
+prost-build = { version = "0.13", optional = true }
+quote = { version = "1.0", optional = true }
+serde = { version = "1.0", features = ["derive"], optional = true }
+serde_json = { version = "1.0", optional = true }
+syn = { version = "2.0", optional = true }

objdiff-core/build.rs

@@ -1,6 +1,11 @@
+#[cfg(feature = "any-arch")]
+mod config_gen;
+
 fn main() {
     #[cfg(feature = "bindings")]
     compile_protos();
+    #[cfg(feature = "any-arch")]
+    config_gen::generate_diff_config();
 }
 
 #[cfg(feature = "bindings")]

objdiff-core/config-schema.json

@@ -0,0 +1,248 @@
+{
+  "properties": [
+    {
+      "id": "functionRelocDiffs",
+      "type": "choice",
+      "default": "name_address",
+      "name": "Function relocation diffs",
+      "description": "How relocation targets will be diffed in the function view.",
+      "items": [
+        {
+          "value": "none",
+          "name": "None"
+        },
+        {
+          "value": "name_address",
+          "name": "Name or address"
+        },
+        {
+          "value": "data_value",
+          "name": "Data value"
+        },
+        {
+          "value": "all",
+          "name": "Name or address, data value"
+        }
+      ]
+    },
+    {
+      "id": "spaceBetweenArgs",
+      "type": "boolean",
+      "default": true,
+      "name": "Space between args",
+      "description": "Adds a space between arguments in the diff output."
+    },
+    {
+      "id": "combineDataSections",
+      "type": "boolean",
+      "default": false,
+      "name": "Combine data sections",
+      "description": "Combines data sections with equal names."
+    },
+    {
+      "id": "arm.archVersion",
+      "type": "choice",
+      "default": "auto",
+      "name": "Architecture version",
+      "description": "ARM architecture version to use for disassembly.",
+      "items": [
+        {
+          "value": "auto",
+          "name": "Auto"
+        },
+        {
+          "value": "v4t",
+          "name": "ARMv4T (GBA)"
+        },
+        {
+          "value": "v5te",
+          "name": "ARMv5TE (DS)"
+        },
+        {
+          "value": "v6k",
+          "name": "ARMv6K (3DS)"
+        }
+      ]
+    },
+    {
+      "id": "arm.unifiedSyntax",
+      "type": "boolean",
+      "default": false,
+      "name": "Unified syntax",
+      "description": "Disassemble as unified assembly language (UAL)."
+    },
+    {
+      "id": "arm.avRegisters",
+      "type": "boolean",
+      "default": false,
+      "name": "Use A/V registers",
+      "description": "Display R0-R3 as A1-A4 and R4-R11 as V1-V8."
+    },
+    {
+      "id": "arm.r9Usage",
+      "type": "choice",
+      "default": "generalPurpose",
+      "name": "Display R9 as",
+      "items": [
+        {
+          "value": "generalPurpose",
+          "name": "R9 or V6",
+          "description": "Use R9 as a general-purpose register."
+        },
+        {
+          "value": "sb",
+          "name": "SB (static base)",
+          "description": "Used for position-independent data (PID)."
+        },
+        {
+          "value": "tr",
+          "name": "TR (TLS register)",
+          "description": "Used for thread-local storage."
+        }
+      ]
+    },
+    {
+      "id": "arm.slUsage",
+      "type": "boolean",
+      "default": false,
+      "name": "Display R10 as SL",
+      "description": "Used for explicit stack limits."
+    },
+    {
+      "id": "arm.fpUsage",
+      "type": "boolean",
+      "default": false,
+      "name": "Display R11 as FP",
+      "description": "Used for frame pointers."
+    },
+    {
+      "id": "arm.ipUsage",
+      "type": "boolean",
+      "default": false,
+      "name": "Display R12 as IP",
+      "description": "Used for interworking and long branches."
+    },
+    {
+      "id": "mips.abi",
+      "type": "choice",
+      "default": "auto",
+      "name": "ABI",
+      "description": "MIPS ABI to use for disassembly.",
+      "items": [
+        {
+          "value": "auto",
+          "name": "Auto"
+        },
+        {
+          "value": "o32",
+          "name": "O32"
+        },
+        {
+          "value": "n32",
+          "name": "N32"
+        },
+        {
+          "value": "n64",
+          "name": "N64"
+        }
+      ]
+    },
+    {
+      "id": "mips.instrCategory",
+      "type": "choice",
+      "default": "auto",
+      "name": "Instruction category",
+      "description": "MIPS instruction category to use for disassembly.",
+      "items": [
+        {
+          "value": "auto",
+          "name": "Auto"
+        },
+        {
+          "value": "cpu",
+          "name": "CPU"
+        },
+        {
+          "value": "rsp",
+          "name": "RSP (N64)"
+        },
+        {
+          "value": "r3000gte",
+          "name": "R3000 GTE (PS1)"
+        },
+        {
+          "value": "r4000allegrex",
+          "name": "R4000 ALLEGREX (PSP)"
+        },
+        {
+          "value": "r5900",
+          "name": "R5900 EE (PS2)"
+        }
+      ]
+    },
+    {
+      "id": "x86.formatter",
+      "type": "choice",
+      "default": "intel",
+      "name": "Format",
+      "description": "x86 disassembly syntax.",
+      "items": [
+        {
+          "value": "intel",
+          "name": "Intel"
+        },
+        {
+          "value": "gas",
+          "name": "AT&T"
+        },
+        {
+          "value": "nasm",
+          "name": "NASM"
+        },
+        {
+          "value": "masm",
+          "name": "MASM"
+        }
+      ]
+    }
+  ],
+  "groups": [
+    {
+      "id": "general",
+      "name": "General",
+      "properties": [
+        "functionRelocDiffs",
+        "spaceBetweenArgs",
+        "combineDataSections"
+      ]
+    },
+    {
+      "id": "arm",
+      "name": "ARM",
+      "properties": [
+        "arm.archVersion",
+        "arm.unifiedSyntax",
+        "arm.avRegisters",
+        "arm.r9Usage",
+        "arm.slUsage",
+        "arm.fpUsage",
+        "arm.ipUsage"
+      ]
+    },
+    {
+      "id": "mips",
+      "name": "MIPS",
+      "properties": [
+        "mips.abi",
+        "mips.instrCategory"
+      ]
+    },
+    {
+      "id": "x86",
+      "name": "x86",
+      "properties": [
+        "x86.formatter"
+      ]
+    }
+  ]
+}

objdiff-core/config_gen.rs

@@ -0,0 +1,491 @@
+use std::{
+    fs::File,
+    path::{Path, PathBuf},
+};
+
+use heck::{ToShoutySnakeCase, ToSnakeCase, ToUpperCamelCase};
+use proc_macro2::TokenStream;
+use quote::{format_ident, quote};
+
+#[derive(Debug, serde::Deserialize)]
+pub struct ConfigSchema {
+    pub properties: Vec<ConfigProperty>,
+    pub groups: Vec<ConfigGroup>,
+}
+
+#[derive(Debug, serde::Deserialize)]
+#[serde(tag = "type")]
+pub enum ConfigProperty {
+    #[serde(rename = "boolean")]
+    Boolean(ConfigPropertyBoolean),
+    #[serde(rename = "choice")]
+    Choice(ConfigPropertyChoice),
+}
+
+#[derive(Debug, serde::Deserialize)]
+pub struct ConfigPropertyBase {
+    pub id: String,
+    pub name: String,
+    pub description: Option<String>,
+}
+
+#[derive(Debug, serde::Deserialize)]
+pub struct ConfigPropertyBoolean {
+    #[serde(flatten)]
+    pub base: ConfigPropertyBase,
+    pub default: bool,
+}
+
+#[derive(Debug, serde::Deserialize)]
+pub struct ConfigPropertyChoice {
+    #[serde(flatten)]
+    pub base: ConfigPropertyBase,
+    pub default: String,
+    pub items: Vec<ConfigPropertyChoiceItem>,
+}
+
+#[derive(Debug, serde::Deserialize)]
+pub struct ConfigPropertyChoiceItem {
+    pub value: String,
+    pub name: String,
+    pub description: Option<String>,
+}
+
+#[derive(Debug, serde::Deserialize)]
+pub struct ConfigGroup {
+    pub id: String,
+    pub name: String,
+    pub description: Option<String>,
+    pub properties: Vec<String>,
+}
+
+fn build_doc(name: &str, description: Option<&str>) -> TokenStream {
+    let mut doc = format!(" {}", name);
+    let mut out = quote! { #[doc = #doc] };
+    if let Some(description) = description {
+        doc = format!(" {}", description);
+        out.extend(quote! { #[doc = ""] });
+        out.extend(quote! { #[doc = #doc] });
+    }
+    out
+}
+
+pub fn generate_diff_config() {
+    let schema_path = Path::new(env!("CARGO_MANIFEST_DIR")).join("config-schema.json");
+    println!("cargo:rerun-if-changed={}", schema_path.display());
+    let schema_file = File::open(schema_path).expect("Failed to open config schema file");
+    let schema: ConfigSchema =
+        serde_json::from_reader(schema_file).expect("Failed to parse config schema");
+
+    let mut enums = TokenStream::new();
+    for property in &schema.properties {
+        let ConfigProperty::Choice(choice) = property else {
+            continue;
+        };
+        let enum_ident = format_ident!("{}", choice.base.id.to_upper_camel_case());
+        let mut variants = TokenStream::new();
+        let mut full_variants = TokenStream::new();
+        let mut variant_info = TokenStream::new();
+        let mut variant_to_str = TokenStream::new();
+        let mut variant_to_name = TokenStream::new();
+        let mut variant_to_description = TokenStream::new();
+        let mut variant_from_str = TokenStream::new();
+        for item in &choice.items {
+            let variant_name = item.value.to_upper_camel_case();
+            let variant_ident = format_ident!("{}", variant_name);
+            let is_default = item.value == choice.default;
+            variants.extend(build_doc(&item.name, item.description.as_deref()));
+            if is_default {
+                variants.extend(quote! { #[default] });
+            }
+            let value = &item.value;
+            variants.extend(quote! {
+                #[serde(rename = #value, alias = #variant_name)]
+                #variant_ident,
+            });
+            full_variants.extend(quote! { #enum_ident::#variant_ident, });
+            variant_to_str.extend(quote! { #enum_ident::#variant_ident => #value, });
+            let name = &item.name;
+            variant_to_name.extend(quote! { #enum_ident::#variant_ident => #name, });
+            if let Some(description) = &item.description {
+                variant_to_description.extend(quote! {
+                    #enum_ident::#variant_ident => Some(#description),
+                });
+            } else {
+                variant_to_description.extend(quote! {
+                    #enum_ident::#variant_ident => None,
+                });
+            }
+            let description = if let Some(description) = &item.description {
+                quote! { Some(#description) }
+            } else {
+                quote! { None }
+            };
+            variant_info.extend(quote! {
+                ConfigEnumVariantInfo {
+                    value: #value,
+                    name: #name,
+                    description: #description,
+                    is_default: #is_default,
+                },
+            });
+            variant_from_str.extend(quote! {
+                if s.eq_ignore_ascii_case(#value) { return Ok(#enum_ident::#variant_ident); }
+            });
+        }
+        enums.extend(quote! {
+            #[derive(Copy, Clone, Debug, Default, Eq, PartialEq, Hash, serde::Deserialize, serde::Serialize)]
+            #[cfg_attr(feature = "wasm", derive(tsify_next::Tsify), tsify(from_wasm_abi))]
+            pub enum #enum_ident {
+                #variants
+            }
+            impl ConfigEnum for #enum_ident {
+                #[inline]
+                fn variants() -> &'static [Self] {
+                    static VARIANTS: &[#enum_ident] = &[#full_variants];
+                    VARIANTS
+                }
+                #[inline]
+                fn variant_info() -> &'static [ConfigEnumVariantInfo] {
+                    static VARIANT_INFO: &[ConfigEnumVariantInfo] = &[
+                        #variant_info
+                    ];
+                    VARIANT_INFO
+                }
+                fn as_str(&self) -> &'static str {
+                    match self {
+                        #variant_to_str
+                    }
+                }
+                fn name(&self) -> &'static str {
+                    match self {
+                        #variant_to_name
+                    }
+                }
+                fn description(&self) -> Option<&'static str> {
+                    match self {
+                        #variant_to_description
+                    }
+                }
+            }
+            impl std::str::FromStr for #enum_ident {
+                type Err = ();
+                fn from_str(s: &str) -> Result<Self, Self::Err> {
+                    #variant_from_str
+                    Err(())
+                }
+            }
+        });
+    }
+
+    let mut groups = TokenStream::new();
+    let mut group_idents = Vec::new();
+    for group in &schema.groups {
+        let ident = format_ident!("CONFIG_GROUP_{}", group.id.to_shouty_snake_case());
+        let id = &group.id;
+        let name = &group.name;
+        let description = if let Some(description) = &group.description {
+            quote! { Some(#description) }
+        } else {
+            quote! { None }
+        };
+        let properties =
+            group.properties.iter().map(|p| format_ident!("{}", p.to_upper_camel_case()));
+        groups.extend(quote! {
+            ConfigPropertyGroup {
+                id: #id,
+                name: #name,
+                description: #description,
+                properties: &[#(ConfigPropertyId::#properties,)*],
+            },
+        });
+        group_idents.push(ident);
+    }
+
+    let mut property_idents = Vec::new();
+    let mut property_variants = TokenStream::new();
+    let mut variant_info = TokenStream::new();
+    let mut config_property_id_to_str = TokenStream::new();
+    let mut config_property_id_to_name = TokenStream::new();
+    let mut config_property_id_to_description = TokenStream::new();
+    let mut config_property_id_to_kind = TokenStream::new();
+    let mut property_fields = TokenStream::new();
+    let mut default_fields = TokenStream::new();
+    let mut get_property_value_variants = TokenStream::new();
+    let mut set_property_value_variants = TokenStream::new();
+    let mut set_property_value_str_variants = TokenStream::new();
+    let mut config_property_id_from_str = TokenStream::new();
+    for property in &schema.properties {
+        let base = match property {
+            ConfigProperty::Boolean(b) => &b.base,
+            ConfigProperty::Choice(c) => &c.base,
+        };
+        let id = &base.id;
+        let enum_ident = format_ident!("{}", id.to_upper_camel_case());
+        property_idents.push(enum_ident.clone());
+        config_property_id_to_str.extend(quote! { Self::#enum_ident => #id, });
+        let name = &base.name;
+        config_property_id_to_name.extend(quote! { Self::#enum_ident => #name, });
+        if let Some(description) = &base.description {
+            config_property_id_to_description.extend(quote! {
+                Self::#enum_ident => Some(#description),
+            });
+        } else {
+            config_property_id_to_description.extend(quote! {
+                Self::#enum_ident => None,
+            });
+        }
+        let doc = build_doc(name, base.description.as_deref());
+        property_variants.extend(quote! { #doc #enum_ident, });
+        property_fields.extend(doc);
+        let field_ident = format_ident!("{}", id.to_snake_case());
+        match property {
+            ConfigProperty::Boolean(b) => {
+                let default = b.default;
+                if default {
+                    property_fields.extend(quote! {
+                        #[serde(default = "default_true")]
+                    });
+                }
+                property_fields.extend(quote! {
+                    pub #field_ident: bool,
+                });
+                default_fields.extend(quote! {
+                    #field_ident: #default,
+                });
+            }
+            ConfigProperty::Choice(_) => {
+                property_fields.extend(quote! {
+                    pub #field_ident: #enum_ident,
+                });
+                default_fields.extend(quote! {
+                    #field_ident: #enum_ident::default(),
+                });
+            }
+        }
+        let property_value = match property {
+            ConfigProperty::Boolean(_) => {
+                quote! { ConfigPropertyValue::Boolean(self.#field_ident) }
+            }
+            ConfigProperty::Choice(_) => {
+                quote! { ConfigPropertyValue::Choice(self.#field_ident.as_str()) }
+            }
+        };
+        get_property_value_variants.extend(quote! {
+            ConfigPropertyId::#enum_ident => #property_value,
+        });
+        match property {
+            ConfigProperty::Boolean(_) => {
+                set_property_value_variants.extend(quote! {
+                    ConfigPropertyId::#enum_ident => {
+                        if let ConfigPropertyValue::Boolean(value) = value {
+                            self.#field_ident = value;
+                            Ok(())
+                        } else {
+                            Err(())
+                        }
+                    },
+                });
+                set_property_value_str_variants.extend(quote! {
+                    ConfigPropertyId::#enum_ident => {
+                        if let Ok(value) = value.parse() {
+                            self.#field_ident = value;
+                            Ok(())
+                        } else {
+                            Err(())
+                        }
+                    },
+                });
+            }
+            ConfigProperty::Choice(_) => {
+                set_property_value_variants.extend(quote! {
+                    ConfigPropertyId::#enum_ident => {
+                        if let ConfigPropertyValue::Choice(value) = value {
+                            if let Ok(value) = value.parse() {
+                                self.#field_ident = value;
+                                Ok(())
+                            } else {
+                                Err(())
+                            }
+                        } else {
+                            Err(())
+                        }
+                    },
+                });
+                set_property_value_str_variants.extend(quote! {
+                    ConfigPropertyId::#enum_ident => {
+                        if let Ok(value) = value.parse() {
+                            self.#field_ident = value;
+                            Ok(())
+                        } else {
+                            Err(())
+                        }
+                    },
+                });
+            }
+        }
+        let description = if let Some(description) = &base.description {
+            quote! { Some(#description) }
+        } else {
+            quote! { None }
+        };
+        variant_info.extend(quote! {
+            ConfigEnumVariantInfo {
+                value: #id,
+                name: #name,
+                description: #description,
+                is_default: false,
+            },
+        });
+        match property {
+            ConfigProperty::Boolean(_) => {
+                config_property_id_to_kind.extend(quote! {
+                    Self::#enum_ident => ConfigPropertyKind::Boolean,
+                });
+            }
+            ConfigProperty::Choice(_) => {
+                config_property_id_to_kind.extend(quote! {
+                    Self::#enum_ident => ConfigPropertyKind::Choice(#enum_ident::variant_info()),
+                });
+            }
+        }
+        let snake_id = id.to_snake_case();
+        config_property_id_from_str.extend(quote! {
+            if s.eq_ignore_ascii_case(#id) || s.eq_ignore_ascii_case(#snake_id) {
+                return Ok(Self::#enum_ident);
+            }
+        });
+    }
+
+    let tokens = quote! {
+        pub trait ConfigEnum: Sized {
+            fn variants() -> &'static [Self];
+            fn variant_info() -> &'static [ConfigEnumVariantInfo];
+            fn as_str(&self) -> &'static str;
+            fn name(&self) -> &'static str;
+            fn description(&self) -> Option<&'static str>;
+        }
+        #[derive(Clone, Debug)]
+        pub struct ConfigEnumVariantInfo {
+            pub value: &'static str,
+            pub name: &'static str,
+            pub description: Option<&'static str>,
+            pub is_default: bool,
+        }
+        #[derive(Copy, Clone, Debug, Eq, PartialEq, Hash)]
+        pub enum ConfigPropertyId {
+            #property_variants
+        }
+        impl ConfigEnum for ConfigPropertyId {
+            #[inline]
+            fn variants() -> &'static [Self] {
+                static VARIANTS: &[ConfigPropertyId] = &[#(ConfigPropertyId::#property_idents,)*];
+                VARIANTS
+            }
+            #[inline]
+            fn variant_info() -> &'static [ConfigEnumVariantInfo] {
+                static VARIANT_INFO: &[ConfigEnumVariantInfo] = &[
+                    #variant_info
+                ];
+                VARIANT_INFO
+            }
+            fn as_str(&self) -> &'static str {
+                match self {
+                    #config_property_id_to_str
+                }
+            }
+            fn name(&self) -> &'static str {
+                match self {
+                    #config_property_id_to_name
+                }
+            }
+            fn description(&self) -> Option<&'static str> {
+                match self {
+                    #config_property_id_to_description
+                }
+            }
+        }
+        impl ConfigPropertyId {
+            pub fn kind(&self) -> ConfigPropertyKind {
+                match self {
+                    #config_property_id_to_kind
+                }
+            }
+        }
+        impl std::str::FromStr for ConfigPropertyId {
+            type Err = ();
+            fn from_str(s: &str) -> Result<Self, Self::Err> {
+                #config_property_id_from_str
+                Err(())
+            }
+        }
+        #[derive(Clone, Debug)]
+        pub struct ConfigPropertyGroup {
+            pub id: &'static str,
+            pub name: &'static str,
+            pub description: Option<&'static str>,
+            pub properties: &'static [ConfigPropertyId],
+        }
+        pub static CONFIG_GROUPS: &[ConfigPropertyGroup] = &[#groups];
+        #[derive(Clone, Debug, Eq, PartialEq, Hash)]
+        pub enum ConfigPropertyValue {
+            Boolean(bool),
+            Choice(&'static str),
+        }
+        impl ConfigPropertyValue {
+            pub fn to_json(&self) -> serde_json::Value {
+                match self {
+                    ConfigPropertyValue::Boolean(value) => serde_json::Value::Bool(*value),
+                    ConfigPropertyValue::Choice(value) => serde_json::Value::String(value.to_string()),
+                }
+            }
+        }
+        #[derive(Clone, Debug)]
+        pub enum ConfigPropertyKind {
+            Boolean,
+            Choice(&'static [ConfigEnumVariantInfo]),
+        }
+        #enums
+        #[inline(always)]
+        fn default_true() -> bool { true }
+        #[derive(Clone, Debug, serde::Deserialize, serde::Serialize)]
+        #[cfg_attr(feature = "wasm", derive(tsify_next::Tsify), tsify(from_wasm_abi))]
+        #[serde(default)]
+        pub struct DiffObjConfig {
+            #property_fields
+        }
+        impl Default for DiffObjConfig {
+            fn default() -> Self {
+                Self {
+                    #default_fields
+                }
+            }
+        }
+        impl DiffObjConfig {
+            pub fn get_property_value(&self, id: ConfigPropertyId) -> ConfigPropertyValue {
+                match id {
+                    #get_property_value_variants
+                }
+            }
+            #[allow(clippy::result_unit_err)]
+            pub fn set_property_value(&mut self, id: ConfigPropertyId, value: ConfigPropertyValue) -> Result<(), ()> {
+                match id {
+                    #set_property_value_variants
+                }
+            }
+            #[allow(clippy::result_unit_err)]
+            pub fn set_property_value_str(&mut self, id: ConfigPropertyId, value: &str) -> Result<(), ()> {
+                match id {
+                    #set_property_value_str_variants
+                }
+            }
+        }
+    };
+    let file = syn::parse2(tokens).unwrap();
+    let formatted = prettyplease::unparse(&file);
+    std::fs::write(
+        PathBuf::from(std::env::var_os("OUT_DIR").unwrap()).join("config.gen.rs"),
+        formatted,
+    )
+    .unwrap();
+}

objdiff-core/protos/diff.proto

@@ -21,9 +21,9 @@ enum SymbolFlag {
   SYMBOL_NONE = 0;
   SYMBOL_GLOBAL = 1;
   SYMBOL_LOCAL = 2;
-  SYMBOL_WEAK = 3;
-  SYMBOL_COMMON = 4;
-  SYMBOL_HIDDEN = 5;
+  SYMBOL_WEAK = 4;
+  SYMBOL_COMMON = 8;
+  SYMBOL_HIDDEN = 16;
 }
 
 // A single parsed instruction
@@ -122,10 +122,17 @@ message InstructionBranchTo {
   uint32 branch_index = 2;
 }
 
-message FunctionDiff {
+message SymbolRef {
+  optional uint32 section_index = 1;
+  uint32 symbol_index = 2;
+}
+
+message SymbolDiff {
   Symbol symbol = 1;
   repeated InstructionDiff instructions = 2;
   optional float match_percent = 3;
+  // The symbol ref in the _other_ object that this symbol was diffed against
+  optional SymbolRef target = 5;
 }
 
 message DataDiff {
@@ -140,7 +147,7 @@ message SectionDiff {
   SectionKind kind = 2;
   uint64 size = 3;
   uint64 address = 4;
-  repeated FunctionDiff functions = 5;
+  repeated SymbolDiff symbols = 5;
   repeated DataDiff data = 6;
   optional float match_percent = 7;
 }

objdiff-core/src/arch/arm.rs

@@ -139,9 +139,9 @@ impl ObjArch for ObjArchArm {
 
         let version = match config.arm_arch_version {
             ArmArchVersion::Auto => self.detected_version.unwrap_or(ArmVersion::V5Te),
-            ArmArchVersion::V4T => ArmVersion::V4T,
-            ArmArchVersion::V5TE => ArmVersion::V5Te,
-            ArmArchVersion::V6K => ArmVersion::V6K,
+            ArmArchVersion::V4t => ArmVersion::V4T,
+            ArmArchVersion::V5te => ArmVersion::V5Te,
+            ArmArchVersion::V6k => ArmVersion::V6K,
         };
         let endian = match self.endianness {
             object::Endianness::Little => unarm::Endian::Little,
@@ -276,6 +276,19 @@ impl ObjArch for ObjArchArm {
     fn display_reloc(&self, flags: RelocationFlags) -> Cow<'static, str> {
         Cow::Owned(format!("<{flags:?}>"))
     }
+
+    fn get_reloc_byte_size(&self, flags: RelocationFlags) -> usize {
+        match flags {
+            RelocationFlags::Elf { r_type } => match r_type {
+                elf::R_ARM_ABS32 => 4,
+                elf::R_ARM_REL32 => 4,
+                elf::R_ARM_ABS16 => 2,
+                elf::R_ARM_ABS8 => 1,
+                _ => 1,
+            },
+            _ => 1,
+        }
+    }
 }
 
 #[derive(Clone, Copy, Debug)]

objdiff-core/src/arch/arm64.rs

@@ -173,6 +173,21 @@ impl ObjArch for ObjArchArm64 {
             _ => Cow::Owned(format!("<{flags:?}>")),
         }
     }
+
+    fn get_reloc_byte_size(&self, flags: RelocationFlags) -> usize {
+        match flags {
+            RelocationFlags::Elf { r_type } => match r_type {
+                elf::R_AARCH64_ABS64 => 8,
+                elf::R_AARCH64_ABS32 => 4,
+                elf::R_AARCH64_ABS16 => 2,
+                elf::R_AARCH64_PREL64 => 8,
+                elf::R_AARCH64_PREL32 => 4,
+                elf::R_AARCH64_PREL16 => 2,
+                _ => 1,
+            },
+            _ => 1,
+        }
+    }
 }
 
 struct DisplayCtx<'a> {

objdiff-core/src/arch/mips.rs

@@ -99,8 +99,8 @@ impl ObjArch for ObjArchMips {
             MipsInstrCategory::Auto => self.instr_category,
             MipsInstrCategory::Cpu => InstrCategory::CPU,
             MipsInstrCategory::Rsp => InstrCategory::RSP,
-            MipsInstrCategory::R3000Gte => InstrCategory::R3000GTE,
-            MipsInstrCategory::R4000Allegrex => InstrCategory::R4000ALLEGREX,
+            MipsInstrCategory::R3000gte => InstrCategory::R3000GTE,
+            MipsInstrCategory::R4000allegrex => InstrCategory::R4000ALLEGREX,
             MipsInstrCategory::R5900 => InstrCategory::R5900,
         };
 
@@ -271,6 +271,17 @@ impl ObjArch for ObjArchMips {
             _ => Cow::Owned(format!("<{flags:?}>")),
         }
     }
+
+    fn get_reloc_byte_size(&self, flags: RelocationFlags) -> usize {
+        match flags {
+            RelocationFlags::Elf { r_type } => match r_type {
+                elf::R_MIPS_16 => 2,
+                elf::R_MIPS_32 => 4,
+                _ => 1,
+            },
+            _ => 1,
+        }
+    }
 }
 
 fn push_reloc(args: &mut Vec<ObjInsArg>, reloc: &ObjReloc) -> Result<()> {

objdiff-core/src/arch/mod.rs

@@ -148,6 +148,8 @@ pub trait ObjArch: Send + Sync {
 
     fn display_reloc(&self, flags: RelocationFlags) -> Cow<'static, str>;
 
+    fn get_reloc_byte_size(&self, flags: RelocationFlags) -> usize;
+
     fn symbol_address(&self, symbol: &Symbol) -> u64 { symbol.address() }
 
     fn guess_data_type(&self, _instruction: &ObjIns) -> Option<DataType> { None }
@@ -156,6 +158,17 @@ pub trait ObjArch: Send + Sync {
         Some(format!("Bytes: {:#x?}", bytes))
     }
 
+    fn display_ins_data(&self, ins: &ObjIns) -> Option<String> {
+        let reloc = ins.reloc.as_ref()?;
+        if reloc.addend >= 0 && reloc.target.bytes.len() > reloc.addend as usize {
+            self.guess_data_type(ins).and_then(|ty| {
+                self.display_data_type(ty, &reloc.target.bytes[reloc.addend as usize..])
+            })
+        } else {
+            None
+        }
+    }
+
     // Downcast methods
     #[cfg(feature = "ppc")]
     fn ppc(&self) -> Option<&ppc::ObjArchPpc> { None }

objdiff-core/src/arch/ppc.rs

@@ -1,6 +1,6 @@
 use std::{
     borrow::Cow,
-    collections::{BTreeMap, HashMap},
+    collections::{BTreeMap, HashMap, HashSet},
 };
 
 use anyhow::{bail, ensure, Result};
@@ -10,7 +10,7 @@ use object::{
     elf, File, Object, ObjectSection, ObjectSymbol, Relocation, RelocationFlags, RelocationTarget,
     Symbol, SymbolKind,
 };
-use ppc750cl::{Argument, InsIter, Opcode, ParsedIns, GPR};
+use ppc750cl::{Argument, Arguments, Ins, InsIter, Opcode, ParsedIns, GPR};
 
 use crate::{
     arch::{DataType, ObjArch, ProcessCodeResult},
@@ -143,6 +143,15 @@ impl ObjArch for ObjArchPpc {
                 }
             }
 
+            if reloc.is_none() {
+                if let Some(fake_pool_reloc) = fake_pool_reloc_for_addr.get(&cur_addr) {
+                    // If this instruction has a fake pool relocation, show it as a fake argument
+                    // at the end of the line.
+                    args.push(ObjInsArg::PlainText(" ".into()));
+                    push_reloc(&mut args, fake_pool_reloc)?;
+                }
+            }
+
             ops.push(ins.op as u16);
             let line = line_info.range(..=cur_addr as u64).last().map(|(_, &b)| b);
             insts.push(ObjIns {
@@ -193,24 +202,23 @@ impl ObjArch for ObjArchPpc {
         }
     }
 
+    fn get_reloc_byte_size(&self, flags: RelocationFlags) -> usize {
+        match flags {
+            RelocationFlags::Elf { r_type } => match r_type {
+                elf::R_PPC_ADDR32 => 4,
+                elf::R_PPC_UADDR32 => 4,
+                _ => 1,
+            },
+            _ => 1,
+        }
+    }
+
     fn guess_data_type(&self, instruction: &ObjIns) -> Option<super::DataType> {
         if instruction.reloc.as_ref().is_some_and(|r| r.target.name.starts_with("@stringBase")) {
             return Some(DataType::String);
         }
 
-        let op = Opcode::from(instruction.op as u8);
-        if let Some(ty) = guess_data_type_from_load_store_inst_op(op) {
-            Some(ty)
-        } else if op == Opcode::Addi {
-            // Assume that any addi instruction that references a local symbol is loading a string.
-            // This hack is not ideal and results in tons of false positives where it will show
-            // garbage strings (e.g. misinterpreting arrays, float literals, etc).
-            // But not all strings are in the @stringBase pool, so the condition above that checks
-            // the target symbol name would miss some.
-            Some(DataType::String)
-        } else {
-            None
-        }
+        guess_data_type_from_load_store_inst_op(Opcode::from(instruction.op as u8))
     }
 
     fn display_data_type(&self, ty: DataType, bytes: &[u8]) -> Option<String> {
@@ -248,6 +256,12 @@ fn push_reloc(args: &mut Vec<ObjInsArg>, reloc: &ObjReloc) -> Result<()> {
             elf::R_PPC_ADDR32 | elf::R_PPC_UADDR32 | elf::R_PPC_REL24 | elf::R_PPC_REL14 => {
                 args.push(ObjInsArg::Reloc);
             }
+            elf::R_PPC_NONE => {
+                // Fake pool relocation.
+                args.push(ObjInsArg::PlainText("<".into()));
+                args.push(ObjInsArg::Reloc);
+                args.push(ObjInsArg::PlainText(">".into()));
+            }
             _ => bail!("Unsupported ELF PPC relocation type {r_type}"),
         },
         flags => bail!("Unsupported PPC relocation kind: {flags:?}"),
@@ -442,16 +456,43 @@ fn get_offset_and_addr_gpr_for_possible_pool_reference(
                 Argument::Simm(simm),
             ) => Some((simm.0, addr_src_gpr, Some(addr_dst_gpr))),
             (
+                // `mr` or `mr.`
                 Opcode::Or,
                 Argument::GPR(addr_dst_gpr),
                 Argument::GPR(addr_src_gpr),
                 Argument::None,
-            ) => Some((0, addr_src_gpr, Some(addr_dst_gpr))), // `mr` or `mr.`
+            ) => Some((0, addr_src_gpr, Some(addr_dst_gpr))),
+            (
+                Opcode::Add,
+                Argument::GPR(addr_dst_gpr),
+                Argument::GPR(addr_src_gpr),
+                Argument::GPR(_offset_gpr),
+            ) => Some((0, addr_src_gpr, Some(addr_dst_gpr))),
             _ => None,
         }
     }
 }
 
+// Remove the relocation we're keeping track of in a particular register when an instruction reuses
+// that register to hold some other value, unrelated to pool relocation addresses.
+fn clear_overwritten_gprs(ins: Ins, gpr_pool_relocs: &mut HashMap<u8, ObjReloc>) {
+    let mut def_args = Arguments::default();
+    ins.parse_defs(&mut def_args);
+    for arg in def_args {
+        if let Argument::GPR(gpr) = arg {
+            if ins.op == Opcode::Lmw {
+                // `lmw` overwrites all registers from rd to r31.
+                // ppc750cl only returns rd itself, so we manually clear the rest of them.
+                for reg in gpr.0..31 {
+                    gpr_pool_relocs.remove(&reg);
+                }
+                break;
+            }
+            gpr_pool_relocs.remove(&gpr.0);
+        }
+    }
+}
+
 // We create a fake relocation for an instruction, vaguely simulating what the actual relocation
 // might have looked like if it wasn't pooled. This is so minimal changes are needed to display
 // pooled accesses vs non-pooled accesses. We set the relocation type to R_PPC_NONE to indicate that
@@ -461,35 +502,51 @@ fn get_offset_and_addr_gpr_for_possible_pool_reference(
 fn make_fake_pool_reloc(offset: i16, cur_addr: u32, pool_reloc: &ObjReloc) -> Option<ObjReloc> {
     let offset_from_pool = pool_reloc.addend + offset as i64;
     let target_address = pool_reloc.target.address.checked_add_signed(offset_from_pool)?;
-    let orig_section_index = pool_reloc.target.orig_section_index?;
-    // We also need to create a fake target symbol to go inside our fake relocation.
-    // This is because we don't have access to list of all symbols in this section, so we can't find
-    // the real symbol yet. Instead we make a placeholder that has the correct `orig_section_index`
-    // and `address` fields, and then later on when this information is displayed to the user, we
-    // can find the real symbol by searching through the object's section's symbols for one that
-    // contains this address.
-    let fake_target_symbol = ObjSymbol {
-        name: "".to_string(),
-        demangled_name: None,
-        address: target_address,
-        section_address: 0,
-        size: 0,
-        size_known: false,
-        kind: Default::default(),
-        flags: Default::default(),
-        orig_section_index: Some(orig_section_index),
-        virtual_address: None,
-        original_index: None,
-        bytes: vec![],
+    let target;
+    let addend;
+    if pool_reloc.target.orig_section_index.is_some() {
+        // If the target symbol is within this current object, then we also need to create a fake
+        // target symbol to go inside our fake relocation. This is because we don't have access to
+        // list of all symbols in this section, so we can't find the real symbol within the pool
+        // based on its address yet. Instead we make a placeholder that has the correct
+        // `orig_section_index` and `address` fields, and then later on when this information is
+        // displayed to the user, we can find the real symbol by searching through the object's
+        // section's symbols for one that contains this address.
+        target = ObjSymbol {
+            name: "".to_string(),
+            demangled_name: None,
+            address: target_address,
+            section_address: 0,
+            size: 0,
+            size_known: false,
+            kind: Default::default(),
+            flags: Default::default(),
+            orig_section_index: pool_reloc.target.orig_section_index,
+            virtual_address: None,
+            original_index: None,
+            bytes: vec![],
+        };
+        // The addend is also fake because we don't know yet if the `target_address` here is the exact
+        // start of the symbol or if it's in the middle of it.
+        addend = 0;
+    } else {
+        // But if the target symbol is in a different object (extern), then we simply copy the pool
+        // relocation's target. This is because it won't be possible to locate the actual symbol
+        // later on based only off of an offset without knowing the object or section it's in. And
+        // doing that for external symbols would also be unnecessary, because when the compiler
+        // generates an instruction that accesses an external "pool" plus some offset, that won't be
+        // a normal pool that contains other symbols within it that we want to display. It will be
+        // something like a vtable for a class with multiple inheritance (for example, dCcD_Cyl in
+        // The Wind Waker). So just showing that vtable symbol plus an addend to represent the
+        // offset into it works fine in this case, no fake symbol to hold an address is necessary.
+        target = pool_reloc.target.clone();
+        addend = pool_reloc.addend;
     };
-    // The addend is also fake because we don't know yet if the `target_address` here is the exact
-    // start of the symbol or if it's in the middle of it.
-    let fake_addend = 0;
     Some(ObjReloc {
         flags: RelocationFlags::Elf { r_type: elf::R_PPC_NONE },
         address: cur_addr as u64,
-        target: fake_target_symbol,
-        addend: fake_addend,
+        target,
+        addend,
     })
 }
 
@@ -497,78 +554,184 @@ fn make_fake_pool_reloc(offset: i16, cur_addr: u32, pool_reloc: &ObjReloc) -> Op
 // of pooled data relocations in them, finding which instructions load data from those addresses,
 // and constructing a mapping of the address of that instruction to a "fake pool relocation" that
 // simulates what that instruction's relocation would look like if data hadn't been pooled.
-// Limitations: This method currently only goes through the instructions in a function in linear
-// order, from start to finish. It does *not* follow any branches. This means that it could have
-// false positives or false negatives in determining which relocation is currently loaded in which
-// register at any given point in the function, as control flow is not respected.
-// There are currently no known examples of this method producing inaccurate results in reality, but
-// if examples are found, it may be possible to update this method to also follow all branches so
-// that it produces more accurate results.
+// This method tries to follow the function's proper control flow. It keeps track of a queue of
+// states it hasn't traversed yet, where each state holds an instruction address and a HashMap of
+// which registers hold which pool relocations at that point.
+// When a conditional or unconditional branch is encountered, the destination of the branch is added
+// to the queue. Conditional branches will traverse both the path where the branch is taken and the
+// one where it's not. Unconditional branches only follow the branch, ignoring any code immediately
+// after the branch instruction.
+// Limitations: This method cannot read jump tables. This is because the jump tables are located in
+// the .data section, but ObjArch.process_code only has access to the .text section. In order to
+// work around this limitation and avoid completely missing most code inside switch statements that
+// use jump tables, we instead guess that any parts of a function we missed were switch cases, and
+// traverse them as if the last `bctr` before that address had branched there. This should be fairly
+// accurate in practice - in testing the only instructions it seems to miss are double branches that
+// the compiler generates in error which can never be reached during normal execution anyway.
 fn generate_fake_pool_reloc_for_addr_mapping(
-    address: u64,
+    func_address: u64,
     code: &[u8],
     relocations: &[ObjReloc],
 ) -> HashMap<u32, ObjReloc> {
-    let mut active_pool_relocs = HashMap::new();
+    let mut visited_ins_addrs = HashSet::new();
     let mut pool_reloc_for_addr = HashMap::new();
-    for (cur_addr, ins) in InsIter::new(code, address as u32) {
-        let simplified = ins.simplified();
-        let reloc = relocations.iter().find(|r| (r.address as u32 & !3) == cur_addr);
+    let mut ins_iters_with_gpr_state =
+        vec![(InsIter::new(code, func_address as u32), HashMap::new())];
+    let mut gpr_state_at_bctr = BTreeMap::new();
+    while let Some((ins_iter, mut gpr_pool_relocs)) = ins_iters_with_gpr_state.pop() {
+        for (cur_addr, ins) in ins_iter {
+            if visited_ins_addrs.contains(&cur_addr) {
+                // Avoid getting stuck in an infinite loop when following looping branches.
+                break;
+            }
+            visited_ins_addrs.insert(cur_addr);
 
-        if let Some(reloc) = reloc {
-            // This instruction has a real relocation, so it may be a pool load we want to keep
-            // track of.
-            let args = &simplified.args;
-            match (ins.op, args[0], args[1], args[2]) {
-                (
-                    Opcode::Addi,
-                    Argument::GPR(addr_dst_gpr),
-                    Argument::GPR(_addr_src_gpr),
-                    Argument::Simm(_simm),
-                ) => {
-                    active_pool_relocs.insert(addr_dst_gpr.0, reloc.clone()); // `lis` + `addi`
+            let simplified = ins.simplified();
+
+            // First handle traversing the function's control flow.
+            let mut branch_dest = None;
+            for arg in simplified.args_iter() {
+                if let Argument::BranchDest(dest) = arg {
+                    let dest = cur_addr.wrapping_add_signed(dest.0);
+                    branch_dest = Some(dest);
+                    break;
                 }
-                (
-                    Opcode::Ori,
-                    Argument::GPR(addr_dst_gpr),
-                    Argument::GPR(_addr_src_gpr),
-                    Argument::Uimm(_uimm),
-                ) => {
-                    active_pool_relocs.insert(addr_dst_gpr.0, reloc.clone()); // `lis` + `ori`
-                }
-                (Opcode::B, _, _, _) => {
-                    if simplified.mnemonic == "bl" {
-                        // When encountering a function call, clear any active pool relocations from
-                        // the volatile registers (r0, r3-r12), but not the nonvolatile registers.
-                        active_pool_relocs.remove(&0);
-                        for gpr in 3..12 {
-                            active_pool_relocs.remove(&gpr);
+            }
+            if let Some(branch_dest) = branch_dest {
+                if branch_dest >= func_address as u32
+                    && (branch_dest - func_address as u32) < code.len() as u32
+                {
+                    let dest_offset_into_func = branch_dest - func_address as u32;
+                    let dest_code_slice = &code[dest_offset_into_func as usize..];
+                    match ins.op {
+                        Opcode::Bc => {
+                            // Conditional branch.
+                            // Add the branch destination to the queue to do later.
+                            ins_iters_with_gpr_state.push((
+                                InsIter::new(dest_code_slice, branch_dest),
+                                gpr_pool_relocs.clone(),
+                            ));
+                            // Then continue on with the current iterator.
                         }
+                        Opcode::B => {
+                            if simplified.mnemonic != "bl" {
+                                // Unconditional branch.
+                                // Add the branch destination to the queue.
+                                ins_iters_with_gpr_state.push((
+                                    InsIter::new(dest_code_slice, branch_dest),
+                                    gpr_pool_relocs.clone(),
+                                ));
+                                // Break out of the current iterator so we can do the newly added one.
+                                break;
+                            }
+                        }
+                        _ => unreachable!(),
                     }
                 }
-                _ => {}
             }
-        } else if let Some((offset, addr_src_gpr, addr_dst_gpr)) =
-            get_offset_and_addr_gpr_for_possible_pool_reference(ins.op, &simplified)
-        {
-            // This instruction doesn't have a real relocation, so it may be a reference to one of
-            // the already-loaded pools.
-            if let Some(pool_reloc) = active_pool_relocs.get(&addr_src_gpr.0) {
-                if let Some(fake_pool_reloc) = make_fake_pool_reloc(offset, cur_addr, pool_reloc) {
-                    pool_reloc_for_addr.insert(cur_addr, fake_pool_reloc);
+            if let Opcode::Bcctr = ins.op {
+                if simplified.mnemonic == "bctr" {
+                    // Unconditional branch to count register.
+                    // Likely a jump table.
+                    gpr_state_at_bctr.insert(cur_addr, gpr_pool_relocs.clone());
                 }
-                if let Some(addr_dst_gpr) = addr_dst_gpr {
-                    // If the address of the pool relocation got copied into another register, we
-                    // need to keep track of it in that register too as future instructions may
-                    // reference the symbol indirectly via this new register, instead of the
-                    // register the symbol's address was originally loaded into.
-                    // For example, the start of the function might `lis` + `addi` the start of the
-                    // ...data pool into r25, and then later the start of a loop will `addi` r25
-                    // with the offset within the .data section of an array variable into r21.
-                    // Then the body of the loop will `lwzx` one of the array elements from r21.
-                    let mut new_reloc = pool_reloc.clone();
-                    new_reloc.addend += offset as i64;
-                    active_pool_relocs.insert(addr_dst_gpr.0, new_reloc);
+            }
+
+            // Then handle keeping track of which GPR contains which pool relocation.
+            let reloc = relocations.iter().find(|r| (r.address as u32 & !3) == cur_addr);
+            if let Some(reloc) = reloc {
+                // This instruction has a real relocation, so it may be a pool load we want to keep
+                // track of.
+                let args = &simplified.args;
+                match (ins.op, args[0], args[1], args[2]) {
+                    (
+                        // `lis` + `addi`
+                        Opcode::Addi,
+                        Argument::GPR(addr_dst_gpr),
+                        Argument::GPR(_addr_src_gpr),
+                        Argument::Simm(_simm),
+                    ) => {
+                        gpr_pool_relocs.insert(addr_dst_gpr.0, reloc.clone());
+                    }
+                    (
+                        // `lis` + `ori`
+                        Opcode::Ori,
+                        Argument::GPR(addr_dst_gpr),
+                        Argument::GPR(_addr_src_gpr),
+                        Argument::Uimm(_uimm),
+                    ) => {
+                        gpr_pool_relocs.insert(addr_dst_gpr.0, reloc.clone());
+                    }
+                    (Opcode::B, _, _, _) => {
+                        if simplified.mnemonic == "bl" {
+                            // When encountering a function call, clear any active pool relocations from
+                            // the volatile registers (r0, r3-r12), but not the nonvolatile registers.
+                            gpr_pool_relocs.remove(&0);
+                            for gpr in 3..12 {
+                                gpr_pool_relocs.remove(&gpr);
+                            }
+                        }
+                    }
+                    _ => {
+                        clear_overwritten_gprs(ins, &mut gpr_pool_relocs);
+                    }
+                }
+            } else if let Some((offset, addr_src_gpr, addr_dst_gpr)) =
+                get_offset_and_addr_gpr_for_possible_pool_reference(ins.op, &simplified)
+            {
+                // This instruction doesn't have a real relocation, so it may be a reference to one of
+                // the already-loaded pools.
+                if let Some(pool_reloc) = gpr_pool_relocs.get(&addr_src_gpr.0) {
+                    if let Some(fake_pool_reloc) =
+                        make_fake_pool_reloc(offset, cur_addr, pool_reloc)
+                    {
+                        pool_reloc_for_addr.insert(cur_addr, fake_pool_reloc);
+                    }
+                    if let Some(addr_dst_gpr) = addr_dst_gpr {
+                        // If the address of the pool relocation got copied into another register, we
+                        // need to keep track of it in that register too as future instructions may
+                        // reference the symbol indirectly via this new register, instead of the
+                        // register the symbol's address was originally loaded into.
+                        // For example, the start of the function might `lis` + `addi` the start of the
+                        // ...data pool into r25, and then later the start of a loop will `addi` r25
+                        // with the offset within the .data section of an array variable into r21.
+                        // Then the body of the loop will `lwzx` one of the array elements from r21.
+                        let mut new_reloc = pool_reloc.clone();
+                        new_reloc.addend += offset as i64;
+                        gpr_pool_relocs.insert(addr_dst_gpr.0, new_reloc);
+                    } else {
+                        clear_overwritten_gprs(ins, &mut gpr_pool_relocs);
+                    }
+                } else {
+                    clear_overwritten_gprs(ins, &mut gpr_pool_relocs);
+                }
+            } else {
+                clear_overwritten_gprs(ins, &mut gpr_pool_relocs);
+            }
+        }
+
+        // Finally, if we're about to finish the outer loop and don't have any more control flow to
+        // follow, we check if there are any instruction addresses in this function that we missed.
+        // If so, and if there were any `bctr` instructions before those points in this function,
+        // then we try to traverse those missing spots as switch cases.
+        if ins_iters_with_gpr_state.is_empty() {
+            let unseen_addrs = (func_address as u32..func_address as u32 + code.len() as u32)
+                .step_by(4)
+                .filter(|addr| !visited_ins_addrs.contains(addr));
+            for unseen_addr in unseen_addrs {
+                let prev_bctr_gpr_state = gpr_state_at_bctr
+                    .iter()
+                    .filter(|(&addr, _)| addr < unseen_addr)
+                    .min_by_key(|(&addr, _)| addr)
+                    .map(|(_, gpr_state)| gpr_state);
+                if let Some(gpr_pool_relocs) = prev_bctr_gpr_state {
+                    let dest_offset_into_func = unseen_addr - func_address as u32;
+                    let dest_code_slice = &code[dest_offset_into_func as usize..];
+                    ins_iters_with_gpr_state.push((
+                        InsIter::new(dest_code_slice, unseen_addr),
+                        gpr_pool_relocs.clone(),
+                    ));
+                    break;
                 }
             }
         }

objdiff-core/src/arch/x86.rs

@@ -162,6 +162,19 @@ impl ObjArch for ObjArchX86 {
             _ => Cow::Owned(format!("<{flags:?}>")),
         }
     }
+
+    fn get_reloc_byte_size(&self, flags: RelocationFlags) -> usize {
+        match flags {
+            RelocationFlags::Coff { typ } => match typ {
+                pe::IMAGE_REL_I386_DIR16 => 2,
+                pe::IMAGE_REL_I386_REL16 => 2,
+                pe::IMAGE_REL_I386_DIR32 => 4,
+                pe::IMAGE_REL_I386_REL32 => 4,
+                _ => 1,
+            },
+            _ => 1,
+        }
+    }
 }
 
 fn replace_arg(

objdiff-core/src/bindings/diff.rs

@@ -4,6 +4,7 @@ use crate::{
         ObjDataDiff, ObjDataDiffKind, ObjDiff, ObjInsArgDiff, ObjInsBranchFrom, ObjInsBranchTo,
         ObjInsDiff, ObjInsDiffKind, ObjSectionDiff, ObjSymbolDiff,
     },
+    obj,
     obj::{
         ObjInfo, ObjIns, ObjInsArg, ObjInsArgValue, ObjReloc, ObjSectionKind, ObjSymbol,
         ObjSymbolFlagSet, ObjSymbolFlags,
@@ -39,14 +40,14 @@ impl ObjectDiff {
 impl SectionDiff {
     pub fn new(obj: &ObjInfo, section_index: usize, section_diff: &ObjSectionDiff) -> Self {
         let section = &obj.sections[section_index];
-        let functions = section_diff.symbols.iter().map(|d| FunctionDiff::new(obj, d)).collect();
+        let symbols = section_diff.symbols.iter().map(|d| SymbolDiff::new(obj, d)).collect();
         let data = section_diff.data_diff.iter().map(|d| DataDiff::new(obj, d)).collect();
         Self {
             name: section.name.to_string(),
             kind: SectionKind::from(section.kind) as i32,
             size: section.size,
             address: section.address,
-            functions,
+            symbols,
             data,
             match_percent: section_diff.match_percent,
         }
@@ -64,13 +65,22 @@ impl From<ObjSectionKind> for SectionKind {
     }
 }
 
-impl FunctionDiff {
+impl From<obj::SymbolRef> for SymbolRef {
+    fn from(value: obj::SymbolRef) -> Self {
+        Self {
+            section_index: if value.section_idx == obj::SECTION_COMMON {
+                None
+            } else {
+                Some(value.section_idx as u32)
+            },
+            symbol_index: value.symbol_idx as u32,
+        }
+    }
+}
+
+impl SymbolDiff {
     pub fn new(object: &ObjInfo, symbol_diff: &ObjSymbolDiff) -> Self {
         let (_section, symbol) = object.section_symbol(symbol_diff.symbol_ref);
-        // let diff_symbol = symbol_diff.diff_symbol.map(|symbol_ref| {
-        //     let (_section, symbol) = object.section_symbol(symbol_ref);
-        //     Symbol::from(symbol)
-        // });
         let instructions = symbol_diff
             .instructions
             .iter()
@@ -78,9 +88,9 @@ impl FunctionDiff {
             .collect();
         Self {
             symbol: Some(Symbol::new(symbol)),
-            // diff_symbol,
             instructions,
             match_percent: symbol_diff.match_percent,
+            target: symbol_diff.target_symbol.map(SymbolRef::from),
         }
     }
 }
@@ -110,7 +120,7 @@ impl Symbol {
 fn symbol_flags(value: ObjSymbolFlagSet) -> u32 {
     let mut flags = 0u32;
     if value.0.contains(ObjSymbolFlags::Global) {
-        flags |= SymbolFlag::SymbolNone as u32;
+        flags |= SymbolFlag::SymbolGlobal as u32;
     }
     if value.0.contains(ObjSymbolFlags::Local) {
         flags |= SymbolFlag::SymbolLocal as u32;