Support IDO .mdebug for line numbers (#264)

* Support IDO .mdebug for line numbers

Resolves #179

* Fix wasm build && cargo clippy --fix

* Resolve review comment

.github/workflows/build.yaml

@@ -36,9 +36,9 @@ jobs:
       - name: Cache Rust workspace
         uses: Swatinem/rust-cache@v2
       - name: Cargo check
-        run: cargo check --all-targets --all-features
+        run: cargo check --all-targets --all-features --workspace
       - name: Cargo clippy
-        run: cargo clippy --all-targets --all-features
+        run: cargo clippy --all-targets --all-features --workspace
 
   fmt:
     name: Format
@@ -92,7 +92,7 @@ jobs:
       - name: Cache Rust workspace
         uses: Swatinem/rust-cache@v2
       - name: Cargo test
-        run: cargo test --release --features all
+        run: cargo test --release --all-features --workspace
 
   build-cli:
     name: Build objdiff-cli
@@ -146,13 +146,14 @@ jobs:
     steps:
       - name: Checkout
         uses: actions/checkout@v4
+      - name: Install uv
+        if: matrix.build == 'zigbuild'
+        uses: astral-sh/setup-uv@v6
       - name: Install cargo-zigbuild
         if: matrix.build == 'zigbuild'
         run: |
-          python3 -m venv .venv
-          . .venv/bin/activate
-          echo PATH=$PATH >> $GITHUB_ENV
-          pip install ziglang==0.13.0.post1 cargo-zigbuild==0.19.8
+          uv tool install cargo-zigbuild==0.20.1 --with-executables-from ziglang==0.15.1
+          echo "CARGO_ZIGBUILD_ZIG_PATH=$(uv tool dir)/cargo-zigbuild/bin/python-zig" >> $GITHUB_ENV
       - name: Setup Rust toolchain
         uses: dtolnay/rust-toolchain@stable
         with:
@@ -213,13 +214,14 @@ jobs:
           sudo apt-get -y install ${{ matrix.packages }}
       - name: Checkout
         uses: actions/checkout@v4
+      - name: Install uv
+        if: matrix.build == 'zigbuild'
+        uses: astral-sh/setup-uv@v6
       - name: Install cargo-zigbuild
         if: matrix.build == 'zigbuild'
         run: |
-          python3 -m venv .venv
-          . .venv/bin/activate
-          echo PATH=$PATH >> $GITHUB_ENV
-          pip install ziglang==0.13.0.post1 cargo-zigbuild==0.19.8
+          uv tool install cargo-zigbuild==0.20.1 --with-executables-from ziglang==0.15.1
+          echo "CARGO_ZIGBUILD_ZIG_PATH=$(uv tool dir)/cargo-zigbuild/bin/python-zig" >> $GITHUB_ENV
       - name: Setup Rust toolchain
         uses: dtolnay/rust-toolchain@stable
         with:

.pre-commit-config.yaml

@@ -24,7 +24,7 @@ repos:
         description: Run cargo clippy on all project files.
         language: system
         entry: cargo
-        args: ["+nightly", "clippy", "--all-targets", "--all-features"]
+        args: ["+nightly", "clippy", "--all-targets", "--all-features", "--workspace"]
         pass_filenames: false
       - id: cargo-deny
         name: cargo deny

AGENTS.md

@@ -0,0 +1,44 @@
+# Repository Guidelines
+
+## Project Structure & Module Organization
+
+- `objdiff-core`: core library with diffing logic and shared components
+- `objdiff-cli`: CLI/TUI with `ratatui`
+- `objdiff-gui`: GUI with `eframe`/`egui`
+- `objdiff-wasm`: WebAssembly bindings
+
+objdiff has three main frontends: GUI, CLI/TUI, and [web](https://github.com/encounter/objdiff-web) (utilizing `objdiff-wasm`).
+
+`objdiff-gui` is the most fully-featured and is the primary target for new features.
+
+`objdiff-cli` has an interactive TUI `diff` mode (powered by `ratatui`) and a non-interactive `report` mode to generate detailed progress reports (e.g. for <https://decomp.dev>).
+
+`objdiff-wasm` is compiled into a [WebAssembly Component](https://component-model.bytecodealliance.org/). The [API](objdiff-wasm/wit/objdiff.wit) is defined using [WIT](https://component-model.bytecodealliance.org/design/wit.html). The web interface is separate, but mirrors the GUI in functionality.
+
+## Build, Test, and Development Commands
+
+Run `cargo build` for a debug build and `cargo build --release` for an optimized build.
+The CLI can be exercised with `cargo run --release --bin objdiff-cli -- --help`.
+
+The WASM build is not included in the workspace by default due to differences in toolchain and target; build it with `npm -C objdiff-wasm run build`. (nightly required)
+In general, do NOT use `--workspace`, as it will include the WASM crate and likely fail. Formatting and linting commands are exempt.
+
+Pre-commit tasks:
+
+- `cargo test` to run the test suite
+- `cargo +nightly fmt --all` to format code (nightly required)
+- `cargo +nightly clippy --all-targets --all-features --workspace -- -D warnings` to lint code (nightly required)
+- `cargo deny check` (`cargo install --locked cargo-deny` if needed) if dependencies change
+
+## Testing Guidelines
+
+Favor focused unit tests adjacent to the code, and integration scenarios in `objdiff-core/tests`.
+Integration tests utilize snapshots with the `insta` crate; run `curl -LsSf https://insta.rs/install.sh | sh` if needed.
+To generate updated snapshots, use `cargo insta test`, review diffs manually, then accept changes with `cargo insta accept` (or simply `mv` to accept specific snapshots).
+
+## Configuration Tips
+
+- `config.schema.json`: JSON schema for the user-facing project config file (`objdiff.json`)
+- `objdiff-core/config-schema.json`: Internal options schema to generate `DiffObjConfig` (see `objdiff-core/config_gen.rs`)
+
+The project configuration (`objdiff.json`) is intended to be generated by the projects' build system. It includes paths to the target (expected) and base (current) object files, along with build commands and file watch patterns. See `README.md` for a summary of key options.

Cargo.toml

@@ -5,18 +5,28 @@ members = [
     "objdiff-gui",
     "objdiff-wasm",
 ]
+default-members = [
+    "objdiff-cli",
+    "objdiff-core",
+    "objdiff-gui",
+    # Exclude objdiff-wasm by default
+]
 resolver = "3"
 
+[workspace.package]
+version = "3.3.0"
+authors = ["Luke Street <luke@street.dev>"]
+edition = "2024"
+license = "MIT OR Apache-2.0"
+repository = "https://github.com/encounter/objdiff"
+rust-version = "1.88"
+
 [profile.release-lto]
 inherits = "release"
 lto = "fat"
 strip = "debuginfo"
 codegen-units = 1
 
-[workspace.package]
-version = "3.0.0"
-authors = ["Luke Street <luke@street.dev>"]
-edition = "2024"
-license = "MIT OR Apache-2.0"
-repository = "https://github.com/encounter/objdiff"
-rust-version = "1.88"
+[profile.release-min]
+inherits = "release-lto"
+opt-level = "z"

README.md

@@ -7,12 +7,14 @@ A local diffing tool for decompilation projects. Inspired by [decomp.me](https:/
 
 Features:
 
-- Compare entire object files: functions and data.
-- Built-in symbol demangling for C++. (CodeWarrior, Itanium & MSVC)
-- Automatic rebuild on source file changes.
-- Project integration via [configuration file](#configuration).
-- Search and filter all of a project's objects and quickly switch.
-- Click to highlight all instances of values and registers.
+- Compare entire object files: functions and data
+- Built-in C++ symbol demangling (GCC, MSVC, CodeWarrior, Itanium)
+- Automatic rebuild on source file changes
+- Project integration via [configuration file](#configuration)
+- Search and filter objects with quick switching
+- Click-to-highlight values and registers
+- Detailed progress reporting (powers [decomp.dev](https://decomp.dev))
+- WebAssembly API, [web interface](https://github.com/encounter/objdiff-web) and [Visual Studio Code extension](https://marketplace.visualstudio.com/items?itemName=decomp-dev.objdiff) (WIP)
 
 Supports:
 
@@ -40,7 +42,7 @@ For Linux and macOS, run `chmod +x objdiff-*` to make the binary executable.
 
 ### CLI
 
-CLI binaries can be found on the [releases page](https://github.com/encounter/objdiff/releases).
+CLI binaries are available on the [releases page](https://github.com/encounter/objdiff/releases).
 
 ## Screenshots
 
@@ -49,33 +51,30 @@ CLI binaries can be found on the [releases page](https://github.com/encounter/ob
 
 ## Usage
 
-objdiff works by comparing two relocatable object files (`.o`). The objects are expected to have the same relative path
-from the "target" and "base" directories.
+objdiff compares two relocatable object files (`.o`). Here's how it works:
 
-For example, if the target ("expected") object is located at `build/asm/MetroTRK/mslsupp.o` and the base ("actual")
-object is located at `build/src/MetroTRK/mslsupp.o`, the following configuration would be used:
+1. **Create an `objdiff.json` configuration file** in your project root (or generate it with your build script).  
+  This file lists **all objects in the project** with their target ("expected") and base ("current") paths.
 
-- Target build directory: `build/asm`
-- Base build directory: `build/src`
-- Object: `MetroTRK/mslsupp.o`
+2. **Load the project** in objdiff.
 
-objdiff will then execute the build system from the project directory to build both objects:
+3. **Select an object** from the sidebar to begin diffing.
 
-```sh
-$ make build/asm/MetroTRK/mslsupp.o # Only if "Build target object" is enabled
-$ make build/src/MetroTRK/mslsupp.o
-```
+4. **objdiff automatically:**
+   - Executes the build system to compile the base object (from current source code)
+   - Compares the two objects and displays the differences
+   - Watches for source file changes and rebuilds when detected
 
-The objects will then be compared and the results will be displayed in the UI.
+The configuration file allows complete flexibility in project structure - your build directories can have any layout as long as the paths are specified correctly.
 
-See [Configuration](#configuration) for more information.
+See [Configuration](#configuration) for setup details.
 
 ## Configuration
 
-While **not required** (most settings can be specified in the UI), projects can add an `objdiff.json` file to configure the tool automatically. The configuration file must be located in
+Projects can add an `objdiff.json` file to configure the tool automatically. The configuration file must be located in
 the root project directory.
 
-If your project has a generator script (e.g. `configure.py`), it's recommended to generate the objdiff configuration
+If your project has a generator script (e.g. `configure.py`), it's highly recommended to generate the objdiff configuration
 file as well. You can then add `objdiff.json` to your `.gitignore` to prevent it from being committed.
 
 ```json
@@ -90,17 +89,23 @@ file as well. You can then add `objdiff.json` to your `.gitignore` to prevent it
   "build_base": true,
   "watch_patterns": [
     "*.c",
+    "*.cc",
     "*.cp",
     "*.cpp",
     "*.cxx",
+    "*.c++",
     "*.h",
+    "*.hh",
     "*.hp",
     "*.hpp",
     "*.hxx",
+    "*.h++",
+    "*.pch",
+    "*.pch++",
+    "*.inc",
     "*.s",
     "*.S",
     "*.asm",
-    "*.inc",
     "*.py",
     "*.yml",
     "*.txt",
@@ -122,78 +127,69 @@ file as well. You can then add `objdiff.json` to your `.gitignore` to prevent it
 
 ### Schema
 
-View [config.schema.json](config.schema.json) for all available options. The below list is a summary of the most important options.
+> [!NOTE]  
+> View [config.schema.json](config.schema.json) for all available options. Below is a summary of the most important options.
 
-`custom_make` _(optional)_: By default, objdiff will use `make` to build the project.  
-If the project uses a different build system (e.g. `ninja`), specify it here.  
-The build command will be `[custom_make] [custom_args] path/to/object.o`.
+#### Build Configuration
 
-`custom_args` _(optional)_: Additional arguments to pass to the build command prior to the object path.
+**`custom_make`** _(optional, default: `"make"`)_  
+If the project uses a different build system (e.g. `ninja`), specify it here. The build command will be `[custom_make] [custom_args] path/to/object.o`.
 
-`build_target`: If true, objdiff will tell the build system to build the target objects before diffing (e.g.
-  `make path/to/target.o`).  
-This is useful if the target objects are not built by default or can change based on project configuration or edits
-to assembly files.  
-Requires the build system to be configured properly.
+**`custom_args`** _(optional)_  
+Additional arguments to pass to the build command prior to the object path.
 
-`build_base`: If true, objdiff will tell the build system to build the base objects before diffing (e.g. `make path/to/base.o`).  
-It's unlikely you'll want to disable this, unless you're using an external tool to rebuild the base object on source file changes.
+**`build_target`**  _(default: `false`)_  
+If true, objdiff will build the target objects before diffing (e.g. `make path/to/target.o`). Useful if target objects are not built by default or can change based on project configuration. Requires proper build system configuration.
 
-`watch_patterns` _(optional)_: A list of glob patterns to watch for changes.
-([Supported syntax](https://docs.rs/globset/latest/globset/#syntax))  
-If any of these files change, objdiff will automatically rebuild the objects and re-compare them.  
-If not specified, objdiff will use the default patterns listed above.
+**`build_base`**  _(default: `true`)_  
+If true, objdiff will build the base objects before diffing (e.g. `make path/to/base.o`). It's unlikely you'll want to disable this unless using an external tool to rebuild the base object.
 
-`ignore_patterns` _(optional)_: A list of glob patterns to explicitly ignore when watching for changes.  
-([Supported syntax](https://docs.rs/globset/latest/globset/#syntax))  
-If not specified, objdiff will use the default patterns listed above.
+#### File Watching
 
-`units` _(optional)_: If specified, objdiff will display a list of objects in the sidebar for easy navigation.
+**`watch_patterns`** _(optional, default: listed above)_  
+A list of glob patterns to watch for changes ([supported syntax](https://docs.rs/globset/latest/globset/#syntax)). When these files change, objdiff automatically rebuilds and re-compares objects.
 
-> `name` _(optional)_: The name of the object in the UI. If not specified, the object's `path` will be used.
->
-> `target_path`: Path to the "target" or "expected" object from the project root.  
-> This object is the **intended result** of the match.
->
-> `base_path`: Path to the "base" or "actual" object from the project root.  
-> This object is built from the **current source code**.
->
-> `metadata.auto_generated` _(optional)_: Hides the object from the object list, but still includes it in reports.
->
-> `metadata.complete` _(optional)_: Marks the object as "complete" (or "linked") in the object list.  
-> This is useful for marking objects that are fully decompiled. A value of `false` will mark the object as "incomplete".
+**`ignore_patterns`** _(optional, default: listed above)_  
+A list of glob patterns to explicitly ignore when watching for changes ([supported syntax](https://docs.rs/globset/latest/globset/#syntax)).
+
+#### Units (Objects)
+
+**`units`** _(optional)_  
+If specified, objdiff displays a list of objects in the sidebar for easy navigation. Each unit contains:
+
+- **`name`** _(optional)_ - The display name in the UI. Defaults to the object's `path`.
+- **`target_path`** _(optional)_ - Path to the "target" or "expected" object (the **intended result**).
+- **`base_path`** _(optional)_ - Path to the "base" or "current" object (built from **current source code**). Omit if there is no source object yet.
+- **`metadata.auto_generated`** _(optional)_ - Hides the object from the sidebar but includes it in progress reports.
+- **`metadata.complete`** _(optional)_ - Marks the object as "complete" (linked) when `true` or "incomplete" when `false`.
 
 ## Building
 
 Install Rust via [rustup](https://rustup.rs).
 
 ```shell
-$ git clone https://github.com/encounter/objdiff.git
-$ cd objdiff
-$ cargo run --release
+git clone https://github.com/encounter/objdiff.git
+cd objdiff
+cargo run --release
 ```
 
-Or using `cargo install`.
+Or install directly with cargo:
 
 ```shell
-$ cargo install --locked --git https://github.com/encounter/objdiff.git objdiff-gui objdiff-cli
+cargo install --locked --git https://github.com/encounter/objdiff.git objdiff-gui objdiff-cli
 ```
 
-The binaries will be installed to `~/.cargo/bin` as `objdiff` and `objdiff-cli`.
+Binaries will be installed to `~/.cargo/bin` as `objdiff` and `objdiff-cli`.
 
-## Installing `pre-commit`
+## Contributing
 
-When contributing, it's recommended to install `pre-commit` to automatically run the linter and formatter before a commit.
-
-[`uv`](https://github.com/astral-sh/uv#installation) is recommended to manage Python version and tools.
-
-Rust nightly is required for `cargo +nightly fmt` and `cargo +nightly clippy`.
+Install `pre-commit` to run linting and formatting automatically:
 
 ```shell
-$ cargo install --locked cargo-deny
-$ rustup toolchain install nightly
-$ uv tool install pre-commit
-$ pre-commit install
+rustup toolchain install nightly  # Required for cargo fmt/clippy
+cargo install --locked cargo-deny # https://github.com/EmbarkStudios/cargo-deny
+uv tool install pre-commit        # https://docs.astral.sh/uv, or use pipx or pip
+pre-commit install
 ```
 
 ## License

config.schema.json

@@ -15,7 +15,7 @@
     },
     "custom_make": {
       "type": "string",
-      "description": "By default, objdiff will use make to build the project.\nIf the project uses a different build system (e.g. ninja), specify it here.\nThe build command will be `[custom_make] [custom_args] path/to/object.o`.",
+      "description": "If the project uses a different build system (e.g. ninja), specify it here.\nThe build command will be `[custom_make] [custom_args] path/to/object.o`.",
       "examples": [
         "make",
         "ninja"
@@ -41,33 +41,39 @@
     },
     "build_target": {
       "type": "boolean",
-      "description": "If true, objdiff will tell the build system to build the target objects before diffing (e.g. `make path/to/target.o`).\nThis is useful if the target objects are not built by default or can change based on project configuration or edits to assembly files.\nRequires the build system to be configured properly.",
+      "description": "If true, objdiff will build the target objects before diffing (e.g. `make path/to/target.o`).\nUseful if target objects are not built by default or can change based on project configuration.\nRequires proper build system configuration.",
       "default": false
     },
     "build_base": {
       "type": "boolean",
-      "description": "If true, objdiff will tell the build system to build the base objects before diffing (e.g. `make path/to/base.o`).\nIt's unlikely you'll want to disable this, unless you're using an external tool to rebuild the base object on source file changes.",
+      "description": "If true, objdiff will build the base objects before diffing (e.g. `make path/to/base.o`).\nIt's unlikely you'll want to disable this unless using an external tool to rebuild the base object.",
       "default": true
     },
     "watch_patterns": {
       "type": "array",
-      "description": "List of glob patterns to watch for changes in the project.\nIf any of these files change, objdiff will automatically rebuild the objects and re-compare them.\nSupported syntax: https://docs.rs/globset/latest/globset/#syntax",
+      "description": "A list of glob patterns to watch for changes.\nWhen these files change, objdiff automatically rebuilds and re-compares objects.\nSupported syntax: https://docs.rs/globset/latest/globset/#syntax",
       "items": {
         "type": "string"
       },
       "default": [
         "*.c",
+        "*.cc",
         "*.cp",
         "*.cpp",
         "*.cxx",
+        "*.c++",
         "*.h",
+        "*.hh",
         "*.hp",
         "*.hpp",
         "*.hxx",
+        "*.h++",
+        "*.pch",
+        "*.pch++",
+        "*.inc",
         "*.s",
         "*.S",
         "*.asm",
-        "*.inc",
         "*.py",
         "*.yml",
         "*.txt",
@@ -76,7 +82,7 @@
     },
     "ignore_patterns": {
       "type": "array",
-      "description": "List of glob patterns to explicitly ignore when watching for changes.\nFiles matching these patterns will not trigger a rebuild.\nSupported syntax: https://docs.rs/globset/latest/globset/#syntax",
+      "description": "A list of glob patterns to explicitly ignore when watching for changes.\nSupported syntax: https://docs.rs/globset/latest/globset/#syntax",
       "items": {
         "type": "string"
       },
@@ -105,6 +111,25 @@
       "items": {
         "$ref": "#/$defs/progress_category"
       }
+    },
+    "options": {
+      "type": "object",
+      "description": "Diff configuration options that should be applied automatically when the project is loaded.",
+      "additionalProperties": {
+        "oneOf": [
+          {
+            "type": "boolean"
+          },
+          {
+            "type": "string"
+          }
+        ]
+      },
+      "examples": [
+        {
+          "demangler": "gnu_legacy"
+        }
+      ]
     }
   },
   "$defs": {
@@ -113,7 +138,7 @@
       "properties": {
         "name": {
           "type": "string",
-          "description": "The name of the object in the UI. If not specified, the object's path will be used."
+          "description": "The display name in the UI. Defaults to the object's path."
         },
         "path": {
           "type": "string",
@@ -122,11 +147,11 @@
         },
         "target_path": {
           "type": "string",
-          "description": "Path to the target object from the project root.\nRequired if path is not specified."
+          "description": "Path to the \"target\" or \"expected\" object (the intended result)."
         },
         "base_path": {
           "type": "string",
-          "description": "Path to the base object from the project root.\nRequired if path is not specified."
+          "description": "Path to the \"base\" or \"current\" object (built from current source code).\nOmit if there is no source object yet."
         },
         "reverse_fn_order": {
           "type": "boolean",
@@ -150,6 +175,20 @@
           "additionalProperties": {
             "type": "string"
           }
+        },
+        "options": {
+          "type": "object",
+          "description": "Diff configuration options that should be applied when this unit is active.",
+          "additionalProperties": {
+            "oneOf": [
+              {
+                "type": "boolean"
+              },
+              {
+                "type": "string"
+              }
+            ]
+          }
         }
       }
     },
@@ -201,7 +240,7 @@
       "properties": {
         "complete": {
           "type": "boolean",
-          "description": "Marks the object as \"complete\" (or \"linked\") in the object list.\nThis is useful for marking objects that are fully decompiled. A value of `false` will mark the object as \"incomplete\"."
+          "description": "Marks the object as \"complete\" (linked) when `true` or \"incomplete\" when `false`."
         },
         "reverse_fn_order": {
           "type": "boolean",
@@ -221,7 +260,7 @@
         },
         "auto_generated": {
           "type": "boolean",
-          "description": "Hides the object from the object list by default, but still includes it in reports."
+          "description": "Hides the object from the sidebar but includes it in progress reports."
         }
       }
     },

deny.toml

@@ -74,6 +74,7 @@ ignore = [
     #"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-2024-0436", reason = "Unmaintained paste crate is an indirect dependency" },
+    { id = "RUSTSEC-2025-0052", reason = "Unmaintained async-std crate is an 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.
@@ -241,8 +242,8 @@ allow-git = []
 [sources.allow-org]
 # github.com organizations to allow git sources for
 github = [
-    "enarx", # flagset
     "encounter",
+    "gimli-rs",  # gimli
 ]
 # gitlab.com organizations to allow git sources for
 gitlab = []

objdiff-cli/src/cmd/diff.rs

@@ -24,7 +24,8 @@ use objdiff_core::{
         watcher::{Watcher, create_watcher},
     },
     config::{
-        ProjectConfig, ProjectObject, ProjectObjectMetadata, build_globset,
+        ProjectConfig, ProjectObject, ProjectObjectMetadata, ProjectOptions, apply_project_options,
+        build_globset,
         path::{check_path_buf, platform_path, platform_path_serde_option},
     },
     diff::{DiffObjConfig, MappingConfig, ObjectDiff},
@@ -77,11 +78,11 @@ pub struct Args {
 }
 
 pub fn run(args: Args) -> Result<()> {
-    let (target_path, base_path, project_config) =
+    let (target_path, base_path, project_config, unit_options) =
         match (&args.target, &args.base, &args.project, &args.unit) {
             (Some(_), Some(_), None, None)
             | (Some(_), None, None, None)
-            | (None, Some(_), None, None) => (args.target.clone(), args.base.clone(), None),
+            | (None, Some(_), None, None) => (args.target.clone(), args.base.clone(), None, None),
             (None, None, p, u) => {
                 let project = match p {
                     Some(project) => project.clone(),
@@ -106,28 +107,32 @@ pub fn run(args: Args) -> Result<()> {
                     .base_dir
                     .as_ref()
                     .map(|p| project.join(p.with_platform_encoding()));
-                let objects = project_config
-                    .units
+                let units = project_config.units.as_deref().unwrap_or_default();
+                let objects = units
                     .iter()
-                    .flatten()
-                    .map(|o| {
-                        ObjectConfig::new(
-                            o,
-                            &project,
-                            target_obj_dir.as_deref(),
-                            base_obj_dir.as_deref(),
+                    .enumerate()
+                    .map(|(idx, o)| {
+                        (
+                            ObjectConfig::new(
+                                o,
+                                &project,
+                                target_obj_dir.as_deref(),
+                                base_obj_dir.as_deref(),
+                            ),
+                            idx,
                         )
                     })
                     .collect::<Vec<_>>();
-                let object = if let Some(u) = u {
+                let (object, unit_idx) = if let Some(u) = u {
                     objects
                         .iter()
-                        .find(|obj| obj.name == *u)
+                        .find(|(obj, _)| obj.name == *u)
+                        .map(|(obj, idx)| (obj, *idx))
                         .ok_or_else(|| anyhow!("Unit not found: {}", u))?
                 } else if let Some(symbol_name) = &args.symbol {
                     let mut idx = None;
                     let mut count = 0usize;
-                    for (i, obj) in objects.iter().enumerate() {
+                    for (i, (obj, unit_idx)) in objects.iter().enumerate() {
                         if obj
                             .target_path
                             .as_deref()
@@ -135,7 +140,7 @@ pub fn run(args: Args) -> Result<()> {
                             .transpose()?
                             .unwrap_or(false)
                         {
-                            idx = Some(i);
+                            idx = Some((i, *unit_idx));
                             count += 1;
                             if count > 1 {
                                 break;
@@ -144,7 +149,7 @@ pub fn run(args: Args) -> Result<()> {
                     }
                     match (count, idx) {
                         (0, None) => bail!("Symbol not found: {}", symbol_name),
-                        (1, Some(i)) => &objects[i],
+                        (1, Some((i, unit_idx))) => (&objects[i].0, unit_idx),
                         (2.., Some(_)) => bail!(
                             "Multiple instances of {} were found, try specifying a unit",
                             symbol_name
@@ -154,18 +159,29 @@ pub fn run(args: Args) -> Result<()> {
                 } else {
                     bail!("Must specify one of: symbol, project and unit, target and base objects")
                 };
+                let unit_options = units.get(unit_idx).and_then(|u| u.options().cloned());
                 let target_path = object.target_path.clone();
                 let base_path = object.base_path.clone();
-                (target_path, base_path, Some(project_config))
+                (target_path, base_path, Some(project_config), unit_options)
             }
             _ => bail!("Either target and base or project and unit must be specified"),
         };
 
-    run_interactive(args, target_path, base_path, project_config)
+    run_interactive(args, target_path, base_path, project_config, unit_options)
 }
 
-fn build_config_from_args(args: &Args) -> Result<(DiffObjConfig, MappingConfig)> {
+fn build_config_from_args(
+    args: &Args,
+    project_config: Option<&ProjectConfig>,
+    unit_options: Option<&ProjectOptions>,
+) -> Result<(DiffObjConfig, MappingConfig)> {
     let mut diff_config = DiffObjConfig::default();
+    if let Some(options) = project_config.and_then(|config| config.options.as_ref()) {
+        apply_project_options(&mut diff_config, options)?;
+    }
+    if let Some(options) = unit_options {
+        apply_project_options(&mut diff_config, options)?;
+    }
     apply_config_args(&mut diff_config, &args.config)?;
     let mut mapping_config = MappingConfig {
         mappings: Default::default(),
@@ -316,11 +332,13 @@ fn run_interactive(
     target_path: Option<Utf8PlatformPathBuf>,
     base_path: Option<Utf8PlatformPathBuf>,
     project_config: Option<ProjectConfig>,
+    unit_options: Option<ProjectOptions>,
 ) -> Result<()> {
     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 (diff_obj_config, mapping_config) =
+        build_config_from_args(&args, project_config.as_ref(), unit_options.as_ref())?;
     let mut state = AppState {
         jobs: Default::default(),
         waker: Default::default(),

objdiff-cli/src/cmd/report.rs

@@ -7,9 +7,9 @@ use objdiff_core::{
         ChangeItem, ChangeItemInfo, ChangeUnit, Changes, ChangesInput, Measures, REPORT_VERSION,
         Report, ReportCategory, ReportItem, ReportItemMetadata, ReportUnit, ReportUnitMetadata,
     },
-    config::path::platform_path,
-    diff, obj,
-    obj::{SectionKind, SymbolFlag},
+    config::{ProjectObject, ProjectOptions, apply_project_options, path::platform_path},
+    diff,
+    obj::{self, SectionKind, SymbolFlag, SymbolKind},
 };
 use prost::Message;
 use rayon::iter::{IntoParallelRefIterator, ParallelIterator};
@@ -83,14 +83,13 @@ pub fn run(args: Args) -> Result<()> {
 }
 
 fn generate(args: GenerateArgs) -> Result<()> {
-    let mut diff_config = diff::DiffObjConfig {
+    let base_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)?;
 
     let output_format = OutputFormat::from_option(args.format.as_deref())?;
     let project_dir = args.project.as_deref().unwrap_or_else(|| Utf8PlatformPath::new("."));
@@ -101,31 +100,44 @@ fn generate(args: GenerateArgs) -> Result<()> {
         Some((Err(err), _)) => bail!("Failed to load project configuration: {}", err),
         None => bail!("No project configuration found"),
     };
-    info!(
-        "Generating report for {} units (using {} threads)",
-        project.units().len(),
-        if args.deduplicate { 1 } else { rayon::current_num_threads() }
-    );
-
     let target_obj_dir =
         project.target_dir.as_ref().map(|p| project_dir.join(p.with_platform_encoding()));
     let base_obj_dir =
         project.base_dir.as_ref().map(|p| project_dir.join(p.with_platform_encoding()));
-    let objects = project
-        .units
+    let project_units = project.units.as_deref().unwrap_or_default();
+    let objects = project_units
         .iter()
-        .flatten()
-        .map(|o| {
-            ObjectConfig::new(o, project_dir, target_obj_dir.as_deref(), base_obj_dir.as_deref())
+        .enumerate()
+        .map(|(idx, o)| {
+            (
+                ObjectConfig::new(
+                    o,
+                    project_dir,
+                    target_obj_dir.as_deref(),
+                    base_obj_dir.as_deref(),
+                ),
+                idx,
+            )
         })
         .collect::<Vec<_>>();
+    info!(
+        "Generating report for {} units (using {} threads)",
+        objects.len(),
+        if args.deduplicate { 1 } else { rayon::current_num_threads() }
+    );
 
     let start = Instant::now();
     let mut units = vec![];
     let mut existing_functions: HashSet<String> = HashSet::new();
     if args.deduplicate {
         // If deduplicating, we need to run single-threaded
-        for object in &objects {
+        for (object, unit_idx) in &objects {
+            let diff_config = build_unit_diff_config(
+                &base_diff_config,
+                project.options.as_ref(),
+                project_units.get(*unit_idx).and_then(ProjectObject::options),
+                &args.config,
+            )?;
             if let Some(unit) = report_object(object, &diff_config, Some(&mut existing_functions))?
             {
                 units.push(unit);
@@ -134,7 +146,15 @@ fn generate(args: GenerateArgs) -> Result<()> {
     } else {
         let vec = objects
             .par_iter()
-            .map(|object| report_object(object, &diff_config, None))
+            .map(|(object, unit_idx)| {
+                let diff_config = build_unit_diff_config(
+                    &base_diff_config,
+                    project.options.as_ref(),
+                    project_units.get(*unit_idx).and_then(ProjectObject::options),
+                    &args.config,
+                )?;
+                report_object(object, &diff_config, None)
+            })
             .collect::<Result<Vec<Option<ReportUnit>>>>()?;
         units = vec.into_iter().flatten().collect();
     }
@@ -156,6 +176,24 @@ fn generate(args: GenerateArgs) -> Result<()> {
     Ok(())
 }
 
+fn build_unit_diff_config(
+    base: &diff::DiffObjConfig,
+    project_options: Option<&ProjectOptions>,
+    unit_options: Option<&ProjectOptions>,
+    cli_args: &[String],
+) -> Result<diff::DiffObjConfig> {
+    let mut diff_config = base.clone();
+    if let Some(options) = project_options {
+        apply_project_options(&mut diff_config, options)?;
+    }
+    if let Some(options) = unit_options {
+        apply_project_options(&mut diff_config, options)?;
+    }
+    // CLI args override project and unit options
+    apply_config_args(&mut diff_config, cli_args)?;
+    Ok(diff_config)
+}
+
 fn report_object(
     object: &ObjectConfig,
     diff_config: &diff::DiffObjConfig,
@@ -177,14 +215,16 @@ fn report_object(
         .target_path
         .as_ref()
         .map(|p| {
-            obj::read::read(p.as_ref(), diff_config).with_context(|| format!("Failed to open {p}"))
+            obj::read::read(p.as_ref(), diff_config, diff::DiffSide::Target)
+                .with_context(|| format!("Failed to open {p}"))
         })
         .transpose()?;
     let base = object
         .base_path
         .as_ref()
         .map(|p| {
-            obj::read::read(p.as_ref(), diff_config).with_context(|| format!("Failed to open {p}"))
+            obj::read::read(p.as_ref(), diff_config, diff::DiffSide::Base)
+                .with_context(|| format!("Failed to open {p}"))
         })
         .transpose()?;
     let result =
@@ -245,6 +285,7 @@ fn report_object(
                 || symbol.size == 0
                 || symbol.flags.contains(SymbolFlag::Hidden)
                 || symbol.flags.contains(SymbolFlag::Ignored)
+                || symbol.kind == SymbolKind::Section
             {
                 continue;
             }

objdiff-core/Cargo.toml

@@ -41,7 +41,8 @@ any-arch = [
     "dep:regex",
     "dep:similar",
     "dep:syn",
-    "dep:encoding_rs"
+    "dep:encoding_rs",
+    "demangler",
 ]
 bindings = [
     "dep:prost",
@@ -88,38 +89,37 @@ std = [
 ]
 mips = [
     "any-arch",
-    "dep:cpp_demangle",
-    "dep:cwdemangle",
     "dep:rabbitizer",
 ]
 ppc = [
     "any-arch",
-    "dep:cwdemangle",
     "dep:cwextab",
     "dep:powerpc",
     "dep:rlwinmdec",
 ]
 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",
 ]
 superh = [
     "any-arch",
 ]
+demangler = [
+    "dep:cpp_demangle",
+    "dep:cwdemangle",
+    "dep:gnuv2_demangle",
+    "dep:msvc-demangler",
+]
 
 [package.metadata.docs.rs]
 features = ["all"]
@@ -150,7 +150,6 @@ gimli = { git = "https://github.com/gimli-rs/gimli", rev = "7335f00e7c39fd501511
 typed-arena = { version = "2.0", default-features = false, optional = true }
 
 # ppc
-cwdemangle = { version = "1.0", optional = true }
 cwextab = { version = "1.1", optional = true }
 powerpc = { version = "0.4", optional = true }
 rlwinmdec = { version = "1.1", optional = true }
@@ -159,9 +158,7 @@ rlwinmdec = { version = "1.1", optional = true }
 rabbitizer = { version = "2.0.0-alpha.4", default-features = false, features = ["all_extensions"], optional = true }
 
 # x86
-cpp_demangle = { version = "0.4", default-features = false, features = ["alloc"], optional = true }
 iced-x86 = { version = "1.21", default-features = false, features = ["decoder", "intel", "gas", "masm", "nasm", "exhaustive_enums", "no_std"], optional = true }
-msvc-demangler = { version = "0.11", optional = true }
 
 # arm
 unarm = { version = "1.9", optional = true }
@@ -179,6 +176,12 @@ tempfile = { version = "3.20", optional = true }
 time = { version = "0.3", optional = true }
 encoding_rs = { version = "0.8.35", optional = true }
 
+# demangler
+cpp_demangle = { version = "0.4", optional = true, default-features = false, features = ["alloc"] }
+cwdemangle = { version = "1.0", optional = true }
+gnuv2_demangle = { version = "0.1.0", optional = true }
+msvc-demangler = { version = "0.11", optional = true }
+
 [target.'cfg(windows)'.dependencies]
 winapi = { version = "0.3", optional = true, features = ["winbase"] }
 

objdiff-core/config-schema.json

@@ -25,6 +25,42 @@
         }
       ]
     },
+    {
+      "id": "demangler",
+      "type": "choice",
+      "default": "auto",
+      "name": "Demangler",
+      "description": "Which demangler should be used to demangle each symbol.",
+      "items": [
+        {
+          "value": "auto",
+          "name": "Auto",
+          "description": "Try to automatically guess the mangling format."
+        },
+        {
+          "value": "none",
+          "name": "None",
+          "description": "Disable demangling."
+        },
+        {
+          "value": "codewarrior",
+          "name": "CodeWarrior"
+        },
+        {
+          "value": "itanium",
+          "name": "Itanium"
+        },
+        {
+          "value": "msvc",
+          "name": "MSVC"
+        },
+        {
+          "value": "gnu_legacy",
+          "name": "GNU g++ (Legacy)",
+          "description": "Use the old GNU mangling ABI. Used up to g++ 2.9.x"
+        }
+      ]
+    },
     {
       "id": "analyzeDataFlow",
       "type": "boolean",
@@ -158,6 +194,10 @@
           "value": "o32",
           "name": "O32"
         },
+        {
+          "value": "o64",
+          "name": "O64"
+        },
         {
           "value": "n32",
           "name": "N32"
@@ -165,6 +205,14 @@
         {
           "value": "n64",
           "name": "N64"
+        },
+        {
+          "value": "eabi32",
+          "name": "eabi32"
+        },
+        {
+          "value": "eabi64",
+          "name": "eabi64"
         }
       ]
     },
@@ -247,6 +295,7 @@
       "name": "General",
       "properties": [
         "functionRelocDiffs",
+        "demangler",
         "spaceBetweenArgs",
         "combineDataSections",
         "combineTextSections"
@@ -268,26 +317,17 @@
     {
       "id": "mips",
       "name": "MIPS",
-      "properties": [
-        "mips.abi",
-        "mips.instrCategory",
-        "mips.registerPrefix"
-      ]
+      "properties": ["mips.abi", "mips.instrCategory", "mips.registerPrefix"]
     },
     {
       "id": "ppc",
       "name": "PowerPC",
-      "properties": [
-        "ppc.calculatePoolRelocations",
-        "analyzeDataFlow"
-      ]
+      "properties": ["ppc.calculatePoolRelocations", "analyzeDataFlow"]
     },
     {
       "id": "x86",
       "name": "x86",
-      "properties": [
-        "x86.formatter"
-      ]
+      "properties": ["x86.formatter"]
     }
   ]
 }

objdiff-core/src/arch/arm.rs

@@ -1,9 +1,4 @@
-use alloc::{
-    collections::BTreeMap,
-    format,
-    string::{String, ToString},
-    vec::Vec,
-};
+use alloc::{collections::BTreeMap, format, string::ToString, vec::Vec};
 
 use anyhow::{Result, bail};
 use arm_attr::{BuildAttrs, enums::CpuArch, tag::Tag};
@@ -11,7 +6,7 @@ use object::{Endian as _, Object as _, ObjectSection as _, ObjectSymbol as _, el
 use unarm::{args, arm, thumb};
 
 use crate::{
-    arch::{Arch, RelocationOverride, RelocationOverrideTarget},
+    arch::{Arch, OPCODE_DATA, OPCODE_INVALID, RelocationOverride, RelocationOverrideTarget},
     diff::{ArmArchVersion, ArmR9Usage, DiffObjConfig, display::InstructionPart},
     obj::{
         InstructionRef, Relocation, RelocationFlags, ResolvedInstructionRef, ResolvedRelocation,
@@ -164,7 +159,7 @@ impl ArchArm {
             }
             _ => bail!("Invalid instruction size {}", ins_ref.size),
         };
-        let (ins, parsed_ins) = if ins_ref.opcode == u16::MAX {
+        let (ins, parsed_ins) = if ins_ref.opcode == OPCODE_DATA {
             let mut args = args::Arguments::default();
             args[0] = args::Argument::UImm(code);
             let mnemonic = if ins_ref.size == 4 { ".word" } else { ".hword" };
@@ -238,7 +233,7 @@ impl Arch for ArchArm {
                 ops.push(InstructionRef {
                     address: address as u64,
                     size: data.len() as u8,
-                    opcode: u16::MAX,
+                    opcode: OPCODE_DATA,
                     branch_dest: None,
                 });
                 break;
@@ -257,7 +252,7 @@ impl Arch for ArchArm {
                     ops.push(InstructionRef {
                         address: address as u64,
                         size: ins_size as u8,
-                        opcode: u16::MAX,
+                        opcode: OPCODE_INVALID,
                         branch_dest: None,
                     });
                     address += ins_size as u32;
@@ -318,7 +313,7 @@ impl Arch for ArchArm {
                     };
                     (opcode, branch_dest)
                 }
-                unarm::ParseMode::Data => (u16::MAX, None),
+                unarm::ParseMode::Data => (OPCODE_DATA, None),
             };
 
             ops.push(InstructionRef {
@@ -409,12 +404,6 @@ impl Arch for ArchArm {
         }
     }
 
-    fn demangle(&self, name: &str) -> Option<String> {
-        cpp_demangle::Symbol::new(name)
-            .ok()
-            .and_then(|s| s.demangle(&cpp_demangle::DemangleOptions::default()).ok())
-    }
-
     fn reloc_name(&self, flags: RelocationFlags) -> Option<&'static str> {
         match flags {
             RelocationFlags::Elf(r_type) => match r_type {
@@ -471,12 +460,16 @@ impl Arch for ArchArm {
         section: &Section,
         mut next_address: u64,
     ) -> Result<u64> {
-        // Trim any trailing 4-byte zeroes from the end (padding)
-        while next_address >= symbol.address + 4
-            && let Some(data) = section.data_range(next_address - 4, 4)
-            && data == [0u8; 4]
+        // TODO: This should probably check the disasm mode and trim accordingly,
+        // but self.disasm_modes isn't populated until post_init, so it needs a refactor.
+
+        // Trim any trailing 2-byte zeroes from the end (padding)
+        while next_address >= symbol.address + 2
+            && let Some(data) = section.data_range(next_address - 2, 2)
+            && data == [0u8; 2]
+            && section.relocation_at(next_address - 2, 2).is_none()
         {
-            next_address -= 4;
+            next_address -= 2;
         }
         Ok(next_address.saturating_sub(symbol.address))
     }

objdiff-core/src/arch/arm64.rs

@@ -1,8 +1,4 @@
-use alloc::{
-    format,
-    string::{String, ToString},
-    vec::Vec,
-};
+use alloc::{format, string::ToString, vec::Vec};
 use core::cmp::Ordering;
 
 use anyhow::Result;
@@ -14,7 +10,7 @@ use yaxpeax_arm::armv8::a64::{
 };
 
 use crate::{
-    arch::Arch,
+    arch::{Arch, OPCODE_INVALID},
     diff::{DiffObjConfig, display::InstructionPart},
     obj::{
         InstructionRef, Relocation, RelocationFlags, ResolvedInstructionRef, ResolvedRelocation,
@@ -60,7 +56,7 @@ impl Arch for ArchArm64 {
                         ops.push(InstructionRef {
                             address,
                             size: 4,
-                            opcode: u16::MAX,
+                            opcode: OPCODE_INVALID,
                             branch_dest: None,
                         });
                         continue;
@@ -87,7 +83,7 @@ impl Arch for ArchArm64 {
         let decoder = InstDecoder::default();
         let mut ins = Instruction::default();
         if decoder.decode_into(&mut ins, &mut reader).is_err() {
-            cb(InstructionPart::opcode("<invalid>", u16::MAX))?;
+            cb(InstructionPart::opcode("<invalid>", OPCODE_INVALID))?;
             return Ok(());
         }
 
@@ -108,12 +104,6 @@ impl Arch for ArchArm64 {
         Ok(())
     }
 
-    fn demangle(&self, name: &str) -> Option<String> {
-        cpp_demangle::Symbol::new(name)
-            .ok()
-            .and_then(|s| s.demangle(&cpp_demangle::DemangleOptions::default()).ok())
-    }
-
     fn reloc_name(&self, flags: RelocationFlags) -> Option<&'static str> {
         match flags {
             RelocationFlags::Elf(r_type) => match r_type {
@@ -2295,7 +2285,7 @@ where Cb: FnMut(InstructionPart<'static>) {
 // Opcode is #[repr(u16)], but the tuple variants negate that, so we have to do this instead.
 const fn opcode_to_u16(opcode: Opcode) -> u16 {
     match opcode {
-        Opcode::Invalid => u16::MAX,
+        Opcode::Invalid => OPCODE_INVALID,
         Opcode::UDF => 0,
         Opcode::MOVN => 1,
         Opcode::MOVK => 2,

objdiff-core/src/arch/mips.rs

@@ -1,6 +1,6 @@
 use alloc::{
     collections::{BTreeMap, BTreeSet},
-    string::{String, ToString},
+    string::ToString,
     vec::Vec,
 };
 
@@ -12,7 +12,7 @@ use rabbitizer::{
 
 use crate::{
     arch::{Arch, RelocationOverride, RelocationOverrideTarget},
-    diff::{DiffObjConfig, MipsAbi, MipsInstrCategory, display::InstructionPart},
+    diff::{DiffObjConfig, DiffSide, MipsAbi, MipsInstrCategory, display::InstructionPart},
     obj::{
         InstructionArg, InstructionArgValue, InstructionRef, Relocation, RelocationFlags,
         ResolvedInstructionRef, ResolvedRelocation, Section, Symbol, SymbolFlag, SymbolFlagSet,
@@ -27,6 +27,7 @@ pub struct ArchMips {
     pub ri_gp_value: i32,
     pub paired_relocations: Vec<BTreeMap<u64, i64>>,
     pub ignored_symbols: BTreeSet<usize>,
+    pub diff_side: DiffSide,
 }
 
 const EF_MIPS_ABI: u32 = 0x0000F000;
@@ -38,15 +39,18 @@ const EF_MIPS_MACH_5900: u32 = 0x00920000;
 const R_MIPS15_S3: u32 = 119;
 
 impl ArchMips {
-    pub fn new(object: &object::File) -> Result<Self> {
+    pub fn new(object: &object::File, diff_side: DiffSide) -> Result<Self> {
         let mut abi = Abi::O32;
         let mut isa_extension = None;
         match object.flags() {
             object::FileFlags::None => {}
             object::FileFlags::Elf { e_flags, .. } => {
                 abi = match e_flags & EF_MIPS_ABI {
-                    elf::EF_MIPS_ABI_O32 | elf::EF_MIPS_ABI_O64 => Abi::O32,
-                    elf::EF_MIPS_ABI_EABI32 | elf::EF_MIPS_ABI_EABI64 => Abi::N32,
+                    elf::EF_MIPS_ABI_O32 => Abi::O32,
+                    elf::EF_MIPS_ABI_O64 if e_flags & elf::EF_MIPS_ABI2 != 0 => Abi::N64,
+                    elf::EF_MIPS_ABI_O64 => Abi::O64,
+                    elf::EF_MIPS_ABI_EABI32 => Abi::EABI32,
+                    elf::EF_MIPS_ABI_EABI64 => Abi::EABI64,
                     _ => {
                         if e_flags & elf::EF_MIPS_ABI2 != 0 {
                             Abi::N32
@@ -124,7 +128,11 @@ impl ArchMips {
             let Ok(name) = obj_symbol.name() else { continue };
             if let Some(prefix) = name.strip_suffix(".NON_MATCHING") {
                 ignored_symbols.insert(obj_symbol.index().0);
-                if let Some(target_symbol) = object.symbol_by_name(prefix) {
+                // Only remove the prefixless symbols if we are on the Base side of the diff,
+                // to allow diffing against target objects that contain `.NON_MATCHING` markers.
+                if diff_side == DiffSide::Base
+                    && let Some(target_symbol) = object.symbol_by_name(prefix)
+                {
                     ignored_symbols.insert(target_symbol.index().0);
                 }
             }
@@ -137,6 +145,7 @@ impl ArchMips {
             ri_gp_value,
             paired_relocations,
             ignored_symbols,
+            diff_side,
         })
     }
 
@@ -164,8 +173,11 @@ impl ArchMips {
         .with_abi(match diff_config.mips_abi {
             MipsAbi::Auto => self.abi,
             MipsAbi::O32 => Abi::O32,
+            MipsAbi::O64 => Abi::O64,
             MipsAbi::N32 => Abi::N32,
             MipsAbi::N64 => Abi::N64,
+            MipsAbi::Eabi32 => Abi::EABI32,
+            MipsAbi::Eabi64 => Abi::EABI64,
         })
     }
 
@@ -292,13 +304,6 @@ impl Arch for ArchMips {
         }
     }
 
-    fn demangle(&self, name: &str) -> Option<String> {
-        cpp_demangle::Symbol::new(name)
-            .ok()
-            .and_then(|s| s.demangle(&cpp_demangle::DemangleOptions::default()).ok())
-            .or_else(|| cwdemangle::demangle(name, &cwdemangle::DemangleOptions::default()))
-    }
-
     fn reloc_name(&self, flags: RelocationFlags) -> Option<&'static str> {
         match flags {
             RelocationFlags::Elf(r_type) => match r_type {
@@ -350,6 +355,7 @@ impl Arch for ArchMips {
         while new_address >= symbol.address + 4
             && let Some(data) = section.data_range(new_address - 4, 4)
             && data == [0u8; 4]
+            && section.relocation_at(next_address - 4, 4).is_none()
         {
             new_address -= 4;
         }

objdiff-core/src/arch/mod.rs

@@ -17,7 +17,7 @@ use object::Endian as _;
 
 use crate::{
     diff::{
-        DiffObjConfig,
+        DiffObjConfig, DiffSide,
         display::{ContextItem, HoverItem, InstructionPart},
     },
     obj::{
@@ -41,6 +41,9 @@ pub mod superh;
 #[cfg(feature = "x86")]
 pub mod x86;
 
+pub const OPCODE_INVALID: u16 = u16::MAX;
+pub const OPCODE_DATA: u16 = u16::MAX - 1;
+
 /// Represents the type of data associated with an instruction
 #[derive(PartialEq)]
 pub enum DataType {
@@ -368,8 +371,6 @@ pub trait Arch: Any + Debug + Send + Sync {
         Ok(None)
     }
 
-    fn demangle(&self, _name: &str) -> Option<String> { None }
-
     fn reloc_name(&self, _flags: RelocationFlags) -> Option<&'static str> { None }
 
     fn data_reloc_size(&self, flags: RelocationFlags) -> usize;
@@ -418,15 +419,18 @@ pub trait Arch: Any + Debug + Send + Sync {
     }
 }
 
-pub fn new_arch(object: &object::File) -> Result<Box<dyn Arch>> {
+pub fn new_arch(object: &object::File, diff_side: DiffSide) -> Result<Box<dyn Arch>> {
     use object::Object as _;
+    // Avoid unused warnings on non-mips builds
+    let _ = diff_side;
+
     Ok(match object.architecture() {
         #[cfg(feature = "ppc")]
         object::Architecture::PowerPc | object::Architecture::PowerPc64 => {
             Box::new(ppc::ArchPpc::new(object)?)
         }
         #[cfg(feature = "mips")]
-        object::Architecture::Mips => Box::new(mips::ArchMips::new(object)?),
+        object::Architecture::Mips => Box::new(mips::ArchMips::new(object, diff_side)?),
         #[cfg(feature = "x86")]
         object::Architecture::I386 | object::Architecture::X86_64 => {
             Box::new(x86::ArchX86::new(object)?)

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

@@ -21,6 +21,7 @@ use crate::{
     obj::{
         FlowAnalysisResult, InstructionRef, Object, Relocation, RelocationFlags,
         ResolvedInstructionRef, ResolvedRelocation, Section, Symbol, SymbolFlag, SymbolFlagSet,
+        SymbolKind,
     },
 };
 
@@ -307,18 +308,6 @@ impl Arch for ArchPpc {
         }
     }
 
-    fn demangle(&self, mut name: &str) -> Option<String> {
-        if name.starts_with('?') {
-            msvc_demangler::demangle(name, msvc_demangler::DemangleFlags::llvm()).ok()
-        } else {
-            name = name.trim_start_matches('.');
-            cpp_demangle::Symbol::new(name)
-                .ok()
-                .and_then(|s| s.demangle(&cpp_demangle::DemangleOptions::default()).ok())
-                .or_else(|| cwdemangle::demangle(name, &cwdemangle::DemangleOptions::default()))
-        }
-    }
-
     fn reloc_name(&self, flags: RelocationFlags) -> Option<&'static str> {
         match flags {
             RelocationFlags::Elf(r_type) => match r_type {
@@ -468,6 +457,7 @@ impl Arch for ArchPpc {
         while next_address >= symbol.address + 4
             && let Some(data) = section.data_range(next_address - 4, 4)
             && data == [0u8; 4]
+            && section.relocation_at(next_address - 4, 4).is_none()
         {
             next_address -= 4;
         }
@@ -675,7 +665,7 @@ fn make_symbol_ref(symbol: &object::Symbol) -> Result<ExtabSymbolRef> {
 #[derive(Debug)]
 struct PoolReference {
     addr_src_gpr: powerpc::GPR,
-    addr_offset: i16,
+    addr_offset: i64,
     addr_dst_gpr: Option<powerpc::GPR>,
 }
 
@@ -683,16 +673,20 @@ struct PoolReference {
 // If so, return information pertaining to where the instruction is getting that address from and
 // what it's doing with the address (e.g. copying it into another register, adding an offset, etc).
 fn get_pool_reference_for_inst(
-    opcode: powerpc::Opcode,
+    ins: powerpc::Ins,
     simplified: &powerpc::ParsedIns,
 ) -> Option<PoolReference> {
     use powerpc::{Argument, Opcode};
     let args = &simplified.args;
-    if flow_analysis::guess_data_type_from_load_store_inst_op(opcode).is_some() {
+    if flow_analysis::guess_data_type_from_load_store_inst_op(ins.op).is_some() {
         match (args[1], args[2]) {
             (Argument::Offset(offset), Argument::GPR(addr_src_gpr)) => {
                 // e.g. lwz. Immediate offset.
-                Some(PoolReference { addr_src_gpr, addr_offset: offset.0, addr_dst_gpr: None })
+                Some(PoolReference {
+                    addr_src_gpr,
+                    addr_offset: offset.0 as i64,
+                    addr_dst_gpr: None,
+                })
             }
             (Argument::GPR(addr_src_gpr), Argument::GPR(_offset_gpr)) => {
                 // e.g. lwzx. The offset is in a register and was likely calculated from an index.
@@ -712,17 +706,51 @@ fn get_pool_reference_for_inst(
         // If either of these match, we also want to return the destination register that the
         // address is being copied into so that we can detect any future references to that new
         // register as well.
-        match (opcode, args[0], args[1], args[2]) {
+        match (ins.op, args[0], args[1], args[2]) {
             (
+                // `addi` or `subi`
                 Opcode::Addi,
                 Argument::GPR(addr_dst_gpr),
                 Argument::GPR(addr_src_gpr),
                 Argument::Simm(simm),
-            ) => Some(PoolReference {
-                addr_src_gpr,
-                addr_offset: simm.0,
-                addr_dst_gpr: Some(addr_dst_gpr),
-            }),
+            ) => {
+                let offset = if simplified.mnemonic == "addi" { simm.0 } else { -simm.0 };
+                Some(PoolReference {
+                    addr_src_gpr,
+                    addr_offset: offset as i64,
+                    addr_dst_gpr: Some(addr_dst_gpr),
+                })
+            }
+            (
+                // `addis`
+                Opcode::Addis,
+                Argument::GPR(addr_dst_gpr),
+                Argument::GPR(addr_src_gpr),
+                Argument::Uimm(uimm), // Note: `addis` uses UIMM, unlike `addi`, `subi`, and `subis`
+            ) => {
+                assert_eq!(simplified.mnemonic, "addis");
+                let offset = (uimm.0 as i64) << 16;
+                Some(PoolReference {
+                    addr_src_gpr,
+                    addr_offset: offset,
+                    addr_dst_gpr: Some(addr_dst_gpr),
+                })
+            }
+            (
+                // `subis`
+                Opcode::Addis,
+                Argument::GPR(addr_dst_gpr),
+                Argument::GPR(addr_src_gpr),
+                Argument::Simm(simm),
+            ) => {
+                assert_eq!(simplified.mnemonic, "subis");
+                let offset = (simm.0 as i64) << 16;
+                Some(PoolReference {
+                    addr_src_gpr,
+                    addr_offset: offset,
+                    addr_dst_gpr: Some(addr_dst_gpr),
+                })
+            }
             (
                 // `mr` or `mr.`
                 Opcode::Or,
@@ -777,13 +805,13 @@ fn clear_overwritten_gprs(ins: powerpc::Ins, gpr_pool_relocs: &mut BTreeMap<u8,
 // Also, if this instruction is accessing the middle of a symbol instead of the start, we add an
 // addend to indicate that.
 fn make_fake_pool_reloc(
-    offset: i16,
+    offset: i64,
     cur_addr: u32,
     pool_reloc: &Relocation,
     symbols: &[Symbol],
 ) -> Option<Relocation> {
     let pool_reloc = resolve_relocation(symbols, pool_reloc);
-    let offset_from_pool = pool_reloc.relocation.addend + offset as i64;
+    let offset_from_pool = pool_reloc.relocation.addend + offset;
     let target_address = pool_reloc.symbol.address.checked_add_signed(offset_from_pool)?;
     let target_symbol;
     let addend;
@@ -794,6 +822,7 @@ fn make_fake_pool_reloc(
                 && s.size > 0
                 && !s.flags.contains(SymbolFlag::Hidden)
                 && !s.flags.contains(SymbolFlag::Ignored)
+                && s.kind != SymbolKind::Section
                 && (s.address..s.address + s.size).contains(&target_address)
         })?;
         addend = target_address.checked_sub(symbols[target_symbol].address)? as i64;
@@ -946,7 +975,7 @@ fn generate_fake_pool_relocations_for_function(
                         clear_overwritten_gprs(ins, &mut gpr_pool_relocs);
                     }
                 }
-            } else if let Some(pool_ref) = get_pool_reference_for_inst(ins.op, &simplified) {
+            } else if let Some(pool_ref) = get_pool_reference_for_inst(ins, &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(&pool_ref.addr_src_gpr.0) {
@@ -965,7 +994,7 @@ fn generate_fake_pool_relocations_for_function(
                         // 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 += pool_ref.addr_offset as i64;
+                        new_reloc.addend += pool_ref.addr_offset;
                         gpr_pool_relocs.insert(addr_dst_gpr.0, new_reloc);
                     } else {
                         clear_overwritten_gprs(ins, &mut gpr_pool_relocs);

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

@@ -1,4 +1,4 @@
-use alloc::{collections::BTreeMap, format, string::String, vec, vec::Vec};
+use alloc::{collections::BTreeMap, format, vec, vec::Vec};
 
 use anyhow::Result;
 use object::elf;
@@ -132,12 +132,6 @@ impl Arch for ArchSuperH {
         Ok(())
     }
 
-    fn demangle(&self, name: &str) -> Option<String> {
-        cpp_demangle::Symbol::new(name)
-            .ok()
-            .and_then(|s| s.demangle(&cpp_demangle::DemangleOptions::default()).ok())
-    }
-
     fn reloc_name(&self, flags: RelocationFlags) -> Option<&'static str> {
         match flags {
             RelocationFlags::Elf(r_type) => match r_type {

objdiff-core/src/arch/x86.rs

@@ -1,4 +1,4 @@
-use alloc::{boxed::Box, format, string::String, vec::Vec};
+use alloc::{boxed::Box, format, vec::Vec};
 use core::cmp::Ordering;
 
 use anyhow::{Context, Result, anyhow, bail};
@@ -9,7 +9,7 @@ use iced_x86::{
 use object::{Endian as _, Object as _, ObjectSection as _, elf, pe};
 
 use crate::{
-    arch::{Arch, RelocationOverride, RelocationOverrideTarget},
+    arch::{Arch, OPCODE_DATA, RelocationOverride, RelocationOverrideTarget},
     diff::{DiffObjConfig, X86Formatter, display::InstructionPart},
     obj::{InstructionRef, Relocation, RelocationFlags, ResolvedInstructionRef, Section, Symbol},
 };
@@ -89,8 +89,6 @@ impl ArchX86 {
     }
 }
 
-const DATA_OPCODE: u16 = u16::MAX - 1;
-
 impl Arch for ArchX86 {
     fn scan_instructions_internal(
         &self,
@@ -121,10 +119,43 @@ impl Arch for ArchX86 {
                             out.push(InstructionRef {
                                 address,
                                 size: size as u8,
-                                opcode: DATA_OPCODE,
+                                opcode: OPCODE_DATA,
                                 branch_dest: None,
                             });
+
                             reloc_iter.next();
+
+                            // support .byte arrays after jump tables (they're typically known as indirect tables)
+
+                            let indirect_array_address = address.wrapping_add(size as u64);
+                            let indirect_array_pos = decoder.position();
+
+                            let max_size = code.len().saturating_sub(indirect_array_pos);
+
+                            let indirect_array_size = reloc_iter
+                                .peek()
+                                .map(|next_reloc| {
+                                    next_reloc.address.saturating_sub(indirect_array_address)
+                                        as usize
+                                })
+                                .unwrap_or(max_size)
+                                .min(max_size);
+
+                            if indirect_array_size > 0 {
+                                for i in 0..indirect_array_size {
+                                    out.push(InstructionRef {
+                                        address: indirect_array_address + i as u64,
+                                        size: 1,
+                                        opcode: OPCODE_DATA,
+                                        branch_dest: None,
+                                    });
+                                }
+                                // move decoder to after the array (there can be multiple jump+indirect tables in one function)
+                                let _ =
+                                    decoder.set_position(indirect_array_pos + indirect_array_size);
+                                decoder.set_ip(indirect_array_address + indirect_array_size as u64);
+                            }
+
                             continue 'outer;
                         }
                     }
@@ -154,13 +185,14 @@ impl Arch for ArchX86 {
         diff_config: &DiffObjConfig,
         cb: &mut dyn FnMut(InstructionPart) -> Result<()>,
     ) -> Result<()> {
-        if resolved.ins_ref.opcode == DATA_OPCODE {
+        if resolved.ins_ref.opcode == OPCODE_DATA {
             let (mnemonic, imm) = match resolved.ins_ref.size {
+                1 => (".byte", resolved.code[0] as u64),
                 2 => (".word", self.endianness.read_u16_bytes(resolved.code.try_into()?) as u64),
                 4 => (".dword", self.endianness.read_u32_bytes(resolved.code.try_into()?) as u64),
                 _ => bail!("Unsupported x86 inline data size {}", resolved.ins_ref.size),
             };
-            cb(InstructionPart::opcode(mnemonic, DATA_OPCODE))?;
+            cb(InstructionPart::opcode(mnemonic, OPCODE_DATA))?;
             if resolved.relocation.is_some() {
                 cb(InstructionPart::reloc())?;
             } else {
@@ -268,16 +300,6 @@ impl Arch for ArchX86 {
         Ok(Some(RelocationOverride { target: RelocationOverrideTarget::Keep, addend }))
     }
 
-    fn demangle(&self, name: &str) -> Option<String> {
-        if name.starts_with('?') {
-            msvc_demangler::demangle(name, msvc_demangler::DemangleFlags::llvm()).ok()
-        } else {
-            cpp_demangle::Symbol::new(name)
-                .ok()
-                .and_then(|s| s.demangle(&cpp_demangle::DemangleOptions::default()).ok())
-        }
-    }
-
     fn reloc_name(&self, flags: RelocationFlags) -> Option<&'static str> {
         match self.arch {
             Architecture::X86 => match flags {
@@ -791,4 +813,33 @@ mod test {
         .unwrap();
         assert_eq!(parts, &[InstructionPart::opcode("call", opcode), InstructionPart::reloc()]);
     }
+
+    #[test]
+    fn test_display_1_byte_inline_data() {
+        let arch = ArchX86 { arch: Architecture::X86, endianness: object::Endianness::Little };
+        let code = [0xAB];
+        let mut parts = Vec::new();
+        arch.display_instruction(
+            ResolvedInstructionRef {
+                ins_ref: InstructionRef {
+                    address: 0x1234,
+                    size: 1,
+                    opcode: OPCODE_DATA,
+                    branch_dest: None,
+                },
+                code: &code,
+                ..Default::default()
+            },
+            &DiffObjConfig::default(),
+            &mut |part| {
+                parts.push(part.into_static());
+                Ok(())
+            },
+        )
+        .unwrap();
+        assert_eq!(parts, &[
+            InstructionPart::opcode(".byte", OPCODE_DATA),
+            InstructionPart::unsigned(0xABu64),
+        ]);
+    }
 }

objdiff-core/src/config/mod.rs

@@ -1,6 +1,7 @@
 pub mod path;
 
 use alloc::{
+    borrow::Cow,
     collections::BTreeMap,
     string::{String, ToString},
     vec::Vec,
@@ -45,6 +46,8 @@ pub struct ProjectConfig {
     pub units: Option<Vec<ProjectObject>>,
     #[cfg_attr(feature = "serde", serde(skip_serializing_if = "Option::is_none"))]
     pub progress_categories: Option<Vec<ProjectProgressCategory>>,
+    #[cfg_attr(feature = "serde", serde(skip_serializing_if = "Option::is_none"))]
+    pub options: Option<ProjectOptions>,
 }
 
 impl ProjectConfig {
@@ -116,6 +119,8 @@ pub struct ProjectObject {
     pub metadata: Option<ProjectObjectMetadata>,
     #[cfg_attr(feature = "serde", serde(skip_serializing_if = "Option::is_none"))]
     pub symbol_mappings: Option<BTreeMap<String, String>>,
+    #[cfg_attr(feature = "serde", serde(skip_serializing_if = "Option::is_none"))]
+    pub options: Option<ProjectOptions>,
 }
 
 #[derive(Default, Clone)]
@@ -143,6 +148,15 @@ pub struct ProjectProgressCategory {
     pub name: String,
 }
 
+pub type ProjectOptions = BTreeMap<String, ProjectOptionValue>;
+
+#[derive(Clone, Debug)]
+#[cfg_attr(feature = "serde", derive(serde::Serialize, serde::Deserialize), serde(untagged))]
+pub enum ProjectOptionValue {
+    Bool(bool),
+    String(String),
+}
+
 impl ProjectObject {
     pub fn name(&self) -> &str {
         if let Some(name) = &self.name {
@@ -179,6 +193,8 @@ impl ProjectObject {
     pub fn auto_generated(&self) -> Option<bool> {
         self.metadata.as_ref().and_then(|m| m.auto_generated)
     }
+
+    pub fn options(&self) -> Option<&ProjectOptions> { self.options.as_ref() }
 }
 
 #[derive(Default, Clone, Eq, PartialEq)]
@@ -204,8 +220,9 @@ pub struct ScratchConfig {
 pub const CONFIG_FILENAMES: [&str; 3] = ["objdiff.json", "objdiff.yml", "objdiff.yaml"];
 
 pub const DEFAULT_WATCH_PATTERNS: &[&str] = &[
-    "*.c", "*.cp", "*.cpp", "*.cxx", "*.h", "*.hp", "*.hpp", "*.hxx", "*.s", "*.S", "*.asm",
-    "*.inc", "*.py", "*.yml", "*.txt", "*.json",
+    "*.c", "*.cc", "*.cp", "*.cpp", "*.cxx", "*.c++", "*.h", "*.hh", "*.hp", "*.hpp", "*.hxx",
+    "*.h++", "*.pch", "*.pch++", "*.inc", "*.s", "*.S", "*.asm", "*.py", "*.yml", "*.txt",
+    "*.json",
 ];
 
 pub const DEFAULT_IGNORE_PATTERNS: &[&str] = &["build/**/*"];
@@ -309,3 +326,47 @@ pub fn build_globset(vec: &[Glob]) -> Result<GlobSet, globset::Error> {
     }
     builder.build()
 }
+
+#[cfg(feature = "any-arch")]
+pub fn apply_project_options(
+    diff_config: &mut crate::diff::DiffObjConfig,
+    options: &ProjectOptions,
+) -> Result<()> {
+    use core::str::FromStr;
+
+    use crate::diff::{ConfigEnum, ConfigPropertyId, ConfigPropertyKind};
+
+    let mut result = Ok(());
+    for (key, value) in options.iter() {
+        let property_id = ConfigPropertyId::from_str(key)
+            .map_err(|()| anyhow!("Invalid configuration property: {key}"))?;
+        let value = match value {
+            ProjectOptionValue::Bool(value) => Cow::Borrowed(if *value { "true" } else { "false" }),
+            ProjectOptionValue::String(value) => Cow::Borrowed(value.as_str()),
+        };
+        if diff_config.set_property_value_str(property_id, &value).is_err() {
+            if result.is_err() {
+                // Already returning an error, skip further errors
+                continue;
+            }
+            let mut expected = String::new();
+            match property_id.kind() {
+                ConfigPropertyKind::Boolean => expected.push_str("true, false"),
+                ConfigPropertyKind::Choice(variants) => {
+                    for (idx, variant) in variants.iter().enumerate() {
+                        if idx > 0 {
+                            expected.push_str(", ");
+                        }
+                        expected.push_str(variant.value);
+                    }
+                }
+            }
+            result = Err(anyhow!(
+                "Invalid value for {}. Expected one of: {}",
+                property_id.name(),
+                expected
+            ));
+        }
+    }
+    result
+}