v1.2.3

Author: nico

.github/workflows/test.yml

@@ -0,0 +1,31 @@
+name: Build and test
+
+on: [push, pull_request]
+
+jobs:
+  build:
+    runs-on: ${{ matrix.os }}
+
+    strategy:
+      matrix:
+        os: [macos-latest, ubuntu-latest, windows-latest]
+
+    steps:
+    - uses: actions/checkout@v3
+
+    - name: Set up Ninja
+      uses: ashutoshvarma/[email protected]
+
+    - uses: ilammy/[email protected]
+
+    - name: CMake
+      run: |
+        cmake -GNinja .
+
+    - name: Build
+      run: |
+        ninja
+
+    - name: Test
+      run: |
+        python3 demumble_test.py

.gitignore

@@ -0,0 +1,18 @@
+.*.sw?
+
+# dist.py outputs:
+buildlinux/
+buildmac/
+buildwin/
+demumble-linux.zip
+demumble-mac.zip
+demumble-win.zip
+
+# build outputs when building as described in README.md:
+.ninja_deps
+.ninja_log
+CMakeCache.txt
+CMakeFiles/
+build.ninja
+cmake_install.cmake
+demumble

CMakeLists.txt

@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 3.2.0 FATAL_ERROR)
+cmake_minimum_required(VERSION 3.5.0 FATAL_ERROR)
 project(demumble CXX)
 
 if (UNIX)
@@ -19,9 +19,17 @@ if (UNIX)
 endif()
 
 if (WIN32)
+  # https://gitlab.kitware.com/cmake/cmake/-/issues/20610
+  string(REGEX REPLACE "/GR" "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
+  string(REGEX REPLACE "/EHsc" "" CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
+  add_definitions(-D_HAS_EXCEPTIONS=0)
+
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /Zc:inline /EHs-c- /GR-")
   add_definitions(-D_CRT_SECURE_NO_WARNINGS)  # The LLVM build sets this.
 
+  # Disable cl.exe warnings that LLVM disables as well.
+  set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} /wd4244 /wd4267")
+
   # This is apparently the simplest way to statically link the CRT in CMake:
   string(TOUPPER "${CMAKE_BUILD_TYPE}" build)
   set(flag_var "CMAKE_CXX_FLAGS_${build}")
@@ -33,9 +41,12 @@ endif()
 include_directories(third_party/llvm/include)
 add_executable(demumble
                demumble.cc
+               third_party/llvm/lib/Demangle/Demangle.cpp
                third_party/llvm/lib/Demangle/ItaniumDemangle.cpp
+               third_party/llvm/lib/Demangle/DLangDemangle.cpp
                third_party/llvm/lib/Demangle/MicrosoftDemangle.cpp
                third_party/llvm/lib/Demangle/MicrosoftDemangleNodes.cpp
+               third_party/llvm/lib/Demangle/RustDemangle.cpp
 )
-set_target_properties(demumble PROPERTIES CXX_STANDARD 11
+set_target_properties(demumble PROPERTIES CXX_STANDARD 17
                                           CXX_STANDARD_REQUIRED ON)

README.md

@@ -1,13 +1,15 @@
 # demumble
 
-`demumble` demangles both Itanium and Visual Studio symbols. It runs on both
-POSIX and Windows.
+`demumble` demangles both Itanium and Visual Studio symbols. It runs on
+both POSIX and Windows.
 
     $ demumble _Z4funcPci
     func(char*, int)
     $ demumble '?Fx_i@@YAHP6AHH@Z@Z'
     int __cdecl Fx_i(int (__cdecl *)(int))
-    
+
+It can also demangle Rust and D symbols.
+
 ## Download
 
 There are prebuilt x64 binaries for Linux, Mac (10.9+), and Windows on the
@@ -19,9 +21,9 @@ It has several nice features that c++filt lacks (and lacks many of c++filt's
 features I never use).
 
 Smart about underscores: C++ symbols have an additional leading underscore on
-OS X. `operator new` is mangled as `_Znw` on Linux but `__Znw` on Mac. OS X's
+macOS. `operator new` is mangled as `_Znw` on Linux but `__Znw` on Mac. macOS's
 c++filt automatically strips one leading underscore, but Linux's c++filt
-doesn't. So if you want to demangle a Linux symbol on OS X, you need to pass
+doesn't. So if you want to demangle a Linux symbol on macOS, you need to pass
 `-n` to tell it to not strip the underscore, and if you want to demangle an OS
 X symbol on Linux you likewise need to pass `-_`. demumble just does the right
 thing:
@@ -75,7 +77,7 @@ Optionally print both mangled and demangled names:
 
 ## Build instructions
 
-Use cmake to build: `cmake -G Ninja && ninja`
+Use cmake to build: `cmake -G Ninja . && ninja`
 
 Run tests after building: `python demumble_test.py`
 

RELEASING

@@ -1,13 +1,13 @@
 Push new release branch:
-1. make sure branches 'master' and 'release' are synced up locally
-2. update demumble.cc with new version (with ".git"), then
+1. make sure branches 'main' and 'release' are synced up locally
+2. update kDemumbleVersion in demumble.cc with new version (with ".git"), then
        git commit -am 'mark this 1.0.0.git'
-3. git checkout release; git merge master
-4. fix version number in src/version.cc (it will likely conflict in the above)
-5. commit, tag, push (don't forget to push --tags)
+3. git checkout release; git merge main
+4. fix version number in src/version.cc (it will conflict in the above)
+5. commit, tag, push (don't forget to push --tags), build binaries
        git commit -am v1.0.0; git push origin release
        git tag v1.0.0; git push --tags
-       # Push the 1.0.0.git change on master too:
-       git checkout master; git push origin master
-6. add binaries to https://github.com/nico/demumble/releases
-   build them with `./dist.py` (on the release branch)
+       ./dist.py  # on the 'release' branch
+       # Push the 1.0.0.git change on main too:
+       git checkout main; git push origin main
+6. add demumble-{linux,mac,win}.zip to https://github.com/nico/demumble/releases

demumble.cc

@@ -6,7 +6,7 @@
 
 #include "llvm/Demangle/Demangle.h"
 
-const char kDemumbleVersion[] = "1.2.2";
+const char kDemumbleVersion[] = "1.2.3";
 
 static int print_help(FILE* out) {
   fprintf(out,
@@ -22,15 +22,19 @@ static int print_help(FILE* out) {
   return out == stdout ? 0 : 1;
 }
 
-static void print_demangled(const char* format, const char* s) {
-  if (char* itanium = llvm::itaniumDemangle(s, NULL, NULL, NULL)) {
-    printf(format, itanium, s);
+static void print_demangled(const char* format, std::string_view s,
+                            size_t* n_used) {
+  if (char* itanium = llvm::itaniumDemangle(s)) {
+    printf(format, itanium, (int)s.size(), s.data());
     free(itanium);
-  } else if (char* ms = llvm::microsoftDemangle(s, NULL, NULL, NULL)) {
-    printf(format, ms, s);
+  } else if (char* rust = llvm::rustDemangle(s)) {
+    printf(format, rust, (int)s.size(), s.data());
+    free(rust);
+  } else if (char* ms = llvm::microsoftDemangle(s, n_used, NULL)) {
+    printf(format, ms, (int)s.size(), s.data());
     free(ms);
   } else {
-    printf("%s", s);
+    printf("%.*s", (int)s.size(), s.data());
   }
 }
 
@@ -39,6 +43,12 @@ static bool is_mangle_char_itanium(char c) {
          (c >= '0' && c <= '9') || c == '_' || c == '$';
 }
 
+static bool is_mangle_char_rust(char c) {
+  // See https://rust-lang.github.io/rfcs/2603-rust-symbol-name-mangling-v0.html.
+  return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
+         (c >= '0' && c <= '9') || c == '_';
+}
+
 static bool is_mangle_char_win(char c) {
   return (c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z') ||
          (c >= '0' && c <= '9') || strchr("?_@$", c);
@@ -53,6 +63,11 @@ static bool is_plausible_itanium_prefix(char* s) {
   return strstr(prefix, "_Z");
 }
 
+static bool is_plausible_rust_prefix(char* s) {
+  // Rust symbols start with "_R".
+  return s[0] == '_' && s[1] == 'R';
+}
+
 static char buf[8192];
 int main(int argc, char* argv[]) {
   enum { kPrintAll, kPrintMatching } print_mode = kPrintAll;
@@ -68,9 +83,9 @@ int main(int argc, char* argv[]) {
       ++argv;
       break;
     } else if (argv[1][0] == '-' && argv[1][1] != '-') {
-      for (int i = 1; i < strlen(argv[1]); ++i)
+      for (size_t i = 1; i < strlen(argv[1]); ++i)
         switch (argv[1][i]) {
-        case 'b': print_format = "\"%s\" (%s)"; break;
+        case 'b': print_format = "\"%s\" (%.*s)"; break;
         case 'h': return print_help(stdout);
         case 'm': print_mode = kPrintMatching; break;
         case 'u': setbuf(stdout, NULL); break;
@@ -87,8 +102,11 @@ int main(int argc, char* argv[]) {
     ++argv;
   }
   for (int i = 1; i < argc; ++i) {
-    print_demangled(print_format, argv[i]);
+    size_t used = strlen(argv[i]);
+    print_demangled(print_format, { argv[i], used }, &used);
     printf("\n");
+    if (used < strlen(argv[i]))
+      printf("  unused suffix: %s\n", argv[i] + used);
   }
   if (argc == 1) {  // Read stdin instead.
     // By default, don't demangle types.  Mangled function names are unlikely
@@ -118,20 +136,21 @@ int main(int argc, char* argv[]) {
         else if (is_plausible_itanium_prefix(cur))
           while (cur + n_sym != end && is_mangle_char_itanium(cur[n_sym]))
             ++n_sym;
+        else if (is_plausible_rust_prefix(cur))
+          while (cur + n_sym != end && is_mangle_char_rust(cur[n_sym]))
+            ++n_sym;
         else {
           if (print_mode == kPrintAll)
             printf("_");
           ++cur;
           continue;
         }
 
-        char tmp = cur[n_sym];
-        cur[n_sym] = '\0';
-        print_demangled(print_format, cur);
+        size_t n_used = n_sym;
+        print_demangled(print_format, { cur, n_sym }, &n_used);
         need_separator = true;
-        cur[n_sym] = tmp;
 
-        cur += n_sym;
+        cur += n_used;
       }
     }
   }

demumble_test.py

@@ -1,15 +1,18 @@
-#!/usr/bin/env python
-from __future__ import print_function
+#!/usr/bin/env python3
 import os, re, subprocess, sys
 
 tests = [
     ('demumble hello', 'hello\n'),
     ('demumble _Z4funcPci _Z1fv', 'func(char*, int)\nf()\n'),
     ('demumble < _Z4funcPci _Z1fv', 'func(char*, int)\nf()\n'),
+    ('demumble _RINvNtC3std3mem8align_ofdE _RNvNvC5mylib3foo3bar',
+     'std::mem::align_of::<f64>\nmylib::foo::bar\n'),
+    ('demumble < _RINvNtC3std3mem8align_ofdE _RNvNvC5mylib3foo3bar',
+     'std::mem::align_of::<f64>\nmylib::foo::bar\n'),
     ('demumble ?Fxi@@YAHP6AHH@Z@Z', 'int __cdecl Fxi(int (__cdecl *)(int))\n'),
     ('demumble ??0S@@QEAA@$$QEAU0@@Z', 'public: __cdecl S::S(struct S &&)\n'),
     ('demumble ??_C@_02PCEFGMJL@hi?$AA@', '"hi"\n'),
-    ('demumble __Znwi', 'operator new(int)\n'),  # Strip extra _ (for OS X)
+    ('demumble __Znwi', 'operator new(int)\n'),  # Strip extra _ (for macOS)
     ('demumble < __Znwi', 'operator new(int)\n'),  # Also from stdin
     ('demumble -m hi _Z1fv ho _Z1gv', 'hi\nf()\nho\ng()\n'),
     ('demumble -m < hi_ho _Z1fv ho _Z1gv ?hm', 'f()\ng()\n?hm\n'),
@@ -40,6 +43,9 @@
     ('demumble < _ZZ3fooiENK3$_0clEi',
      'foo(int)::$_0::operator()(int) const\n'),
     ('demumble .?AVNet@@', "class Net `RTTI Type Descriptor Name'\n"),
+    ('demumble < asdf?x@@3HAjkl', 'asdfint xjkl\n'),
+    ('demumble < asdf?x@@3Hjkl', 'asdf?x@@3Hjkl\n'),
+    ('demumble ?x@@3HAjkl', 'int x\n  unused suffix: jkl\n'),
 ]
 
 status = 0
@@ -50,12 +56,12 @@
     if '<' in cmd:
         p = subprocess.Popen(cmd[:cmd.index('<')], stdin=subprocess.PIPE,
                              stdout=subprocess.PIPE, stderr=subprocess.STDOUT,
-                             universal_newlines=True)
+                             encoding='utf-8')
         out = p.communicate(input='\n'.join(cmd[cmd.index('<') + 1:]) + '\n')[0]
     else:
-        out = subprocess.check_output(cmd, universal_newlines=True)
+        out = subprocess.check_output(cmd, encoding='utf-8')
     if (out != t[1] if isinstance(t[1], str) else not t[1].match(out)):
-        print("`%s`: Expected '%s', got '%s'" % (t[0], t[1], out))
+        print(f"`{t[0]}`: Expected '{t[1]}', got '{out}'")
         status = 1
-print("passed" if status == 0 else "failed")
+print('passed' if status == 0 else 'failed')
 sys.exit(status)