C++ Demangler Rewrite

.clang-format

@@ -16,6 +16,8 @@ AllowShortIfStatementsOnASingleLine: false
 AllowShortCaseLabelsOnASingleLine: true
 AllowShortFunctionsOnASingleLine: Inline
 AllowShortLoopsOnASingleLine: false
+InsertBraces: true
+SeparateDefinitionBlocks: Always
 AlignAfterOpenBracket: DontAlign
 AlignEscapedNewlines: DontAlign
 AlignConsecutiveMacros: true

.gitignore

@@ -1,2 +1,8 @@
 build*/
 .cache/
+debug_demangler
+debug_demangler.c
+debug_demangler.dSYM/
+compile_commands.json
+steps-grammars
+**/*.dot

.reuse/dep5

@@ -7,6 +7,10 @@ Files: .github/**
 Copyright: 2021 RizinOrg <info@rizin.re>
 License: LGPL-3.0-only
 
+Files: **/.clang-format
+Copyright: 2025 RizinOrg <info@rizin.re>
+License: LGPL-3.0-only
+
 Files: .clang-format
 Copyright: 2020 RizinOrg <info@rizin.re>
 License: LGPL-3.0-only

README.md

@@ -32,3 +32,68 @@ demangle -s pascal 'OUTPUT_$$_init'
 meson --prefix=/usr -Dinstall_lib=true build
 ninja -C build install
 ```
+
+## Demangler
+
+## Graphviz Tracing
+
+The C++ demangler now supports visual tracing via Graphviz for debugging complex parsing scenarios. Instead of text-based trace logs, the system generates `.dot` files that can be rendered as visual graphs showing the parsing tree.
+
+### Features
+
+- **Visual parsing tree**: Each rule invocation becomes a node in the graph
+- **Success/failure tracking**: Nodes are color-coded:
+  - 🟢 **Green**: Successful rule matches
+  - 🔴 **Red**: Failed rule matches  
+  - 🟠 **Orange**: Backtracked attempts (dashed border)
+  - 🔵 **Blue**: Currently running rules
+- **Position tracking**: Shows parsing position and input snippets
+- **Result display**: Successful nodes show their output
+- **Parent-child relationships**: Edges show rule call hierarchy
+- **Backtracking visualization**: Clearly shows which paths were abandoned
+
+### Usage
+
+Enable tracing by setting the `DEMANGLE_TRACE` environment variable:
+
+```bash
+export DEMANGLE_TRACE=1
+./your_demangler _ZN4llvm8DenseMapIPKN5clang4DeclEPNS_6WeakVHEE
+```
+
+Or compile with `-DENABLE_GRAPHVIZ_TRACE` to always enable tracing.
+
+This will generate a file like `demangle_trace__ZN4llvm8DenseMapIPKN5clang4DeclEPNS_6WeakVHE.dot`.
+
+### Generating Images
+
+Convert the `.dot` file to a visual format:
+
+```bash
+# PNG image
+dot -Tpng demangle_trace.dot -o parsing_tree.png
+
+# SVG (scalable)
+dot -Tsvg demangle_trace.dot -o parsing_tree.svg
+
+# PDF
+dot -Tpdf demangle_trace.dot -o parsing_tree.pdf
+```
+
+### Graph Structure
+
+The generated graph shows:
+- **Root node**: The initial `mangled_name` rule
+- **Child nodes**: All rules called during parsing
+- **Edge colors**: Match the node status (green for success, red for failure, etc.)
+- **Node labels**: Include rule name, position, input snippet, and result
+- **Legend**: Explains the color coding
+
+This visual representation makes it much easier to understand complex backtracking scenarios and identify where parsing issues occur.
+
+## Building
+
+```
+meson -Dbuildtype=debugoptimized -Db_sanitize=address,undefined build
+ninja -C build test
+```

meson.build

@@ -1,104 +1,178 @@
-project('libdemangle', 'c', meson_version: '>=0.49.0')
+project('libdemangle', 'c', meson_version : '>=0.49.0')
 
 cc = meson.get_compiler('c')
 
 common_c_args = []
 libdemangle_c_args = []
 libdemangle_src = [
-  'src' / 'cxx' / 'borland.c',
-  'src' / 'cxx.c',
-  'src' / 'demangler.c',
-  'src' / 'demangler_util.c',
-  'src' / 'java.c',
-  'src' / 'microsoft_demangle.c',
-  'src' / 'msvc.c',
-  'src' / 'objc.c',
-  'src' / 'pascal' / 'pascal.c',
-  'src' / 'rust' / 'punycode.c',
-  'src' / 'rust' / 'rust.c',
-  'src' / 'rust' / 'rust_legacy.c',
-  'src' / 'rust' / 'rust_v0.c',
+    'src' / 'borland.c',
+    'src' / 'cxx.c',
+
+    'src' / 'cplusplus' / 'demangle.c',
+    'src' / 'cplusplus' / 'common.c',
+    'src' / 'cplusplus' / 'param.c',
+    'src' / 'cplusplus' / 'meta.c',
+    'src' / 'cplusplus' / 'parser_combinator.c',
+    'src' / 'cplusplus' / 'ast.c',
+    'src' / 'cplusplus' / 'dot_graph.c',
+    'src' / 'cplusplus' / 'v2.c',
+    'src' / 'cplusplus' / 'v3.c',
+
+    'src' / 'demangler.c',
+    'src' / 'demangler_util.c',
+    'src' / 'java.c',
+    'src' / 'microsoft_demangle.c',
+    'src' / 'msvc.c',
+    'src' / 'objc.c',
+    'src' / 'pascal' / 'pascal.c',
+    'src' / 'rust' / 'punycode.c',
+    'src' / 'rust' / 'rust.c',
+    'src' / 'rust' / 'rust_legacy.c',
+    'src' / 'rust' / 'rust_v0.c',
 ]
 
 tests = [
-  'borland',
-  'java',
-  'msvc',
-  'objc',
-  'pascal',
-  'rust',
+    'borland',
+    'java',
+    'msvc',
+    'objc',
+    'pascal',
+    'rust',
+
+    'subs', # Substitution table tests
+
+    'cxx.00',
+    'cxx.01',
+    'cxx.02',
+    'cxx.03',
+    'cxx.04',
+    'cxx.05',
+    'cxx.06',
+    'cxx.07',
+    'cxx.08',
+    'cxx.09',
+    'cxx.10',
+    'cxx.11',
+    'cxx.12',
+    'cxx.13',
+    'cxx.14',
+    'cxx.15',
+    'cxx.16',
+    'cxx.17',
+    'cxx.18',
+    'cxx.19',
+    'cxx.20',
+    'cxx.21',
+    'cxx.22',
+    'cxx.23',
+    'cxx.24',
+    'cxx.25',
+    'cxx.26',
+
+    'cxx_base',
+    'cxx_gnu_v2',
+    'cxx_gnu_v3_type',
+    'cxx_gnu_v3_template',
+    'cxx_gnu_v3_llvm',
 ]
 
-if get_option('use_gpl')
-  libdemangle_src += 'src' / 'cxx' / 'cp-demangle.c'
-  libdemangle_src += 'src' / 'gnu_v2' / 'cplus-dem.c'
-  common_c_args += '-DWITH_GPL=1'
-  tests += 'cxx'
-  tests += 'cxx_base'
-  tests += 'cxx_gnu_v2'
-endif
+unit_tests = [
+    'cxx_rules',
+    'vec_impl'
+]
 
 if get_option('use_swift_demangler')
-  libdemangle_src += 'src' / 'swift.c'
-  common_c_args += '-DWITH_SWIFT_DEMANGLER=1'
-  tests += 'swift'
+    libdemangle_src += 'src' / 'swift.c'
+    common_c_args += '-DWITH_SWIFT_DEMANGLER=1'
+    tests += 'swift'
 endif
 
 if get_option('default_library') == 'shared'
-  if cc.has_argument('-fvisibility=hidden')
-    libdemangle_c_args += '-fvisibility=hidden'
-  endif
+    if cc.has_argument('-fvisibility=hidden')
+        libdemangle_c_args += '-fvisibility=hidden'
+    endif
 endif
 
 if cc.has_argument('--std=gnu99')
-  add_project_arguments('--std=gnu99', language: 'c')
+    add_project_arguments('--std=gnu99', language : 'c')
 elif cc.has_argument('--std=c99')
-  add_project_arguments('--std=c99', language: 'c')
+    add_project_arguments('--std=c99', language : 'c')
 endif
 
-libdemangle = library('demangle', libdemangle_src,
+libdemangle = both_libraries(
+    'demangle',
+    libdemangle_src,
     c_args : common_c_args + libdemangle_c_args,
-    dependencies: [],
-    implicit_include_directories: false,
-    install: get_option('install_lib'),
-    include_directories: include_directories(['include', 'src']))
+    dependencies : [],
+    implicit_include_directories : false,
+    install : get_option('install_lib'),
+    include_directories : include_directories(['include', 'src']),
+)
 
 libdemangle_dep = declare_dependency(
-    link_with: libdemangle,
-    dependencies: [],
-    include_directories: include_directories('include'),
+    link_with : libdemangle,
+    dependencies : [],
+    include_directories : include_directories('include'),
 )
 
 if get_option('enable_cli')
-  bin_demangle = [
-    'bin' / 'demangle.c',
-  ]
-  executable('demangle', bin_demangle,
-    c_args : common_c_args,
-    dependencies: [libdemangle_dep],
-    include_directories: include_directories(['include', 'src']),
-    implicit_include_directories: false,
-    install: true,
-  )
+    bin_demangle = [
+        'bin' / 'demangle.c',
+    ]
+    executable(
+        'demangle',
+        bin_demangle,
+        c_args : common_c_args,
+        dependencies : [libdemangle_dep],
+        include_directories : include_directories(['include', 'src']),
+        implicit_include_directories : false,
+        install : true,
+    )
 endif
 
 if get_option('enable_tests') and not meson.is_subproject()
-  test_dependencies = [libdemangle_dep]
-  if get_option('b_sanitize').contains('address')
-    test_dependencies += cc.find_library('asan')
-  endif
-  if get_option('b_sanitize').contains('undefined')
-    test_dependencies += cc.find_library('ubsan')
-  endif
-  foreach test : tests
-    exe = executable('test_@0@'.format(test), 'test/test_@0@.c'.format(test),
-      include_directories: include_directories(['include', 'test']),
-      dependencies: test_dependencies,
-      c_args : common_c_args,
-      install: false,
-      install_rpath: '',
-      implicit_include_directories: false
-    )
-    test(test, exe)
-  endforeach
+    test_dependencies = []
+    if get_option('b_sanitize').contains('address')
+        test_dependencies += cc.find_library('asan')
+    endif
+    if get_option('b_sanitize').contains('undefined')
+        test_dependencies += cc.find_library('ubsan')
+    endif
+    # Tests that need longer timeout due to large number of test cases
+    # (especially when running with sanitizers like ASAN)
+    slow_tests = ['cxx_base']
+
+    foreach test : tests
+        exe = executable(
+            'test_@0@'.format(test),
+            'test/test_@0@.c'.format(test),
+            link_whole : libdemangle.get_static_lib(),
+            include_directories : include_directories(['include', 'test']),
+            dependencies : test_dependencies,
+            c_args : common_c_args,
+            install : false,
+            install_rpath : '',
+            implicit_include_directories : false,
+        )
+        if test in slow_tests
+            test(test, exe, timeout : 120)
+        else
+            test(test, exe)
+        endif
+    endforeach
+
+    foreach test : unit_tests
+        exe = executable(
+            'test_@0@'.format(test),
+            'test/unit/test_@0@.c'.format(test),
+            link_whole : libdemangle.get_static_lib(),
+            include_directories : include_directories(['include', 'test']),
+            dependencies : test_dependencies,
+            c_args : common_c_args,
+            install : false,
+            install_rpath : '',
+            implicit_include_directories : false,
+        )
+        test(test, exe)
+    endforeach
 endif

meson_options.txt

@@ -1,5 +1,4 @@
-option('use_gpl', type: 'boolean', value: true, description: 'Set to false when you want to disable gpl code')
 option('use_swift_demangler', type: 'boolean', value: true, description: 'If false, disables the swift demangler')
 option('install_lib', type: 'boolean', value: false, description: 'install libdemangle in the specified prefix path.')
 option('enable_cli', type: 'boolean', value: false, description: 'install a cli to demangle symbols.')
-option('enable_tests', type: 'boolean', value: true, description: 'Build unit tests in test/')
+option('enable_tests', type: 'boolean', value: true, description: 'Build unit tests in test/')