[pull] main from rui314:main

.github/workflows/ci.yml

@@ -1,9 +1,7 @@
 name: CI
 on:
   push:
-    branches: [ main ]
   pull_request:
-    branches: [ main ]
 env:
   UBSAN_OPTIONS: print_stacktrace=1:halt_on_error=1
 jobs:
@@ -13,7 +11,6 @@ jobs:
         target:
         # Disable PCH for the default configuration. This prevents relying on implicit includes.
         - '-DCMAKE_DISABLE_PRECOMPILE_HEADERS=On'
-        - '-DMOLD_USE_ASAN=On'
         - '-DMOLD_USE_TSAN=On'
     runs-on: ubuntu-20.04
     steps:
@@ -72,21 +69,22 @@ jobs:
 
         # Install a LoongArch toolchain
         mkdir /larch
-        wget -O- -q https://github.com/loongson/build-tools/releases/download/2023.08.08/CLFS-loongarch64-8.1-x86_64-cross-tools-gcc-glibc.tar.xz | tar -C /larch --strip-components=1 --xz -xf -
+        wget -O- -q https://github.com/loongson/build-tools/releases/download/2024.06.01/x86_64-cross-tools-loongarch64-binutils_2.42-gcc_14.1.0-glibc_2.39.tar.xz | tar -C /larch --strip-components=1 --xz -xf -
 
+        cp -r /larch/loongarch64-unknown-linux-gnu/lib/* /larch/target/lib64
         ln -sf /larch/target /usr/loongarch64-linux-gnu
-        cp -r /larch/loongarch64-unknown-linux-gnu/lib/* /usr/loongarch64-linux-gnu/lib64/
 
-        for i in objdump objcopy strip; do
+        for i in gcc g++ objdump objcopy strip; do
           ln -sf /larch/bin/loongarch64-unknown-linux-gnu-$i /usr/bin/loongarch64-linux-gnu-$i
         done
 
-        echo '/larch/bin/loongarch64-unknown-linux-gnu-gcc -L/larch/loongarch64-unknown-linux-gnu "$@"' > /usr/bin/loongarch64-linux-gnu-gcc
-        echo '/larch/bin/loongarch64-unknown-linux-gnu-g++ -L/larch/loongarch64-unknown-linux-gnu "$@"' > /usr/bin/loongarch64-linux-gnu-g++
-        chmod 755 /usr/bin/loongarch64-linux-gnu-{gcc,g++}
-
         wget -O /usr/local/bin/qemu-loongarch64 -q https://github.com/loongson/build-tools/releases/download/2023.08.08/qemu-loongarch64
         chmod 755 /usr/local/bin/qemu-loongarch64
+
+        # Install Intel SDE CPU emulator for CET-related tests
+        mkdir /sde
+        wget -O- -q https://downloadmirror.intel.com/813591/sde-external-9.33.0-2024-01-07-lin.tar.xz | tar -C /sde --strip-components=1 --xz -xf -
+        ln -s /sde/sde /usr/bin
     - name: ccache
       uses: hendrikmuhs/ccache-action@v1
     - name: build
@@ -109,8 +107,33 @@ jobs:
           build
           !build/CMakeFiles
 
+  build-distros:
+    strategy:
+      matrix:
+        distro:
+        - alpine
+        - archlinux
+        - fedora
+        - gentoo/stage3
+        - opensuse/tumbleweed
+    runs-on: ubuntu-latest
+    container: ${{ matrix.distro }}
+    steps:
+    - uses: actions/checkout@v2
+    - run: ./install-build-deps.sh
+    - name: build
+      run: |
+        mkdir build
+        cd build
+        cmake ..
+        cmake --build . -j$(nproc)
+    - name: test
+      run: |
+        cd build
+        ctest -j$(nproc)
+
   build-macos:
-    runs-on: macos-11
+    runs-on: macos-12
     strategy:
       matrix:
         target:
@@ -139,3 +162,21 @@ jobs:
         cd build
         cmake -T clangcl ..
         cmake --build . -j $Env:NUMBER_OF_PROCESSORS
+
+  build-msys:
+    runs-on: windows-latest
+    steps:
+    - uses: actions/checkout@v3
+    - name: Setup MSYS2
+      uses: msys2/setup-msys2@v2
+      with:
+        msystem: UCRT64
+        update: true
+        pacboy: gcc-libs:p libwinpthread-git:p tbb:p zlib:p zstd:p dlfcn:p cc:p cmake:p ninja:p
+    - name: build
+      shell: msys2 {0}
+      run: |
+        mkdir build
+        cd build
+        cmake -GNinja -DMOLD_USE_MIMALLOC=OFF -DMOLD_USE_SYSTEM_TBB=ON -DMOLD_USE_MOLD=OFF ..
+        cmake --build . -j $(nproc)

.github/workflows/update-manpage.yml

@@ -1,5 +1,3 @@
-# This file is generated by ChatGPT
-
 name: Update manpage
 
 on:
@@ -8,6 +6,7 @@ on:
       - 'docs/mold.md'
     branches:
       - main
+  workflow_dispatch:
 
 jobs:
   update-manpage:
@@ -18,7 +17,7 @@ jobs:
       uses: actions/checkout@v2
 
     - name: Install ronn
-      run: sudo apt-get install -y ronn
+      run: sudo apt-get update && sudo apt-get install -y ronn
 
     - name: Generate mold.1 from mold.md
       run: ronn --roff docs/mold.md

CMakeLists.txt

@@ -44,7 +44,7 @@
 # features and behave exactly the same.
 
 cmake_minimum_required(VERSION 3.14)
-project(mold VERSION 2.31.0)
+project(mold VERSION 2.32.1)
 
 include(CMakeDependentOption)
 include(CheckSymbolExists)
@@ -58,7 +58,12 @@ endif()
 
 add_executable(mold)
 target_compile_features(mold PRIVATE cxx_std_20)
-target_link_libraries(mold PRIVATE ${CMAKE_DL_LIBS})
+
+if(MINGW)
+  target_link_libraries(mold PRIVATE dl)
+else()
+  target_link_libraries(mold PRIVATE ${CMAKE_DL_LIBS})
+endif()
 
 if(NOT "${CMAKE_CXX_COMPILER_FRONTEND_VARIANT}" STREQUAL "MSVC")
   target_compile_options(mold PRIVATE
@@ -317,17 +322,23 @@ list(APPEND MOLD_ELF_TEMPLATE_FILES
   elf/output-chunks.cc
   elf/passes.cc
   elf/relocatable.cc
-  elf/subprocess.cc
+  elf/shrink-sections.cc
   elf/thunks.cc
   elf/tls.cc
   )
 
-if(WIN32)
+if(WIN32 AND NOT MINGW)
   list(APPEND MOLD_ELF_TEMPLATE_FILES elf/lto-win32.cc)
 else()
   list(APPEND MOLD_ELF_TEMPLATE_FILES elf/lto-unix.cc)
 endif()
 
+if(WIN32)
+  list(APPEND MOLD_ELF_TEMPLATE_FILES elf/subprocess-win32.cc)
+else()
+  list(APPEND MOLD_ELF_TEMPLATE_FILES elf/subprocess-unix.cc)
+endif()
+
 function(mold_instantiate_templates SOURCE TARGET)
   set(PATH ${CMAKE_BINARY_DIR}/${SOURCE}.${TARGET}.cc)
   if(NOT EXISTS ${PATH})
@@ -348,6 +359,7 @@ endforeach()
 # Add other non-template source files.
 target_sources(mold PRIVATE
   common/compress.cc
+  common/crc32.cc
   common/demangle.cc
   common/filepath.cc
   common/glob.cc

common/common.h

@@ -2,6 +2,7 @@
 
 #include "integers.h"
 
+#include <array>
 #include <atomic>
 #include <bit>
 #include <bitset>
@@ -404,9 +405,9 @@ void update_maximum(std::atomic<T> &atomic, u64 new_val, Compare cmp = {}) {
                                        std::memory_order_relaxed));
 }
 
-template <typename T, typename U>
-inline void append(std::vector<T> &vec1, std::vector<U> vec2) {
-  vec1.insert(vec1.end(), vec2.begin(), vec2.end());
+template <typename T>
+inline void append(std::vector<T> &x, const auto &y) {
+  x.insert(x.end(), y.begin(), y.end());
 }
 
 template <typename T>
@@ -534,14 +535,6 @@ std::string_view save_string(Context &ctx, const std::string &str) {
   return {(char *)buf, str.size()};
 }
 
-inline bool remove_prefix(std::string_view &s, std::string_view prefix) {
-  if (s.starts_with(prefix)) {
-    s = s.substr(prefix.size());
-    return true;
-  }
-  return false;
-}
-
 static inline void pause() {
 #if defined(__x86_64__)
   asm volatile("pause");
@@ -734,15 +727,15 @@ template <typename Context>
 class OutputFile {
 public:
   static std::unique_ptr<OutputFile<Context>>
-  open(Context &ctx, std::string path, i64 filesize, i64 perm);
+  open(Context &ctx, std::string path, i64 filesize, int perm);
 
   virtual void close(Context &ctx) = 0;
   virtual ~OutputFile() = default;
 
   u8 *buf = nullptr;
   std::vector<u8> buf2;
   std::string path;
-  i64 fd = -1;
+  int fd = -1;
   i64 filesize = 0;
   bool is_mmapped = false;
   bool is_unmapped = false;
@@ -755,7 +748,7 @@ class OutputFile {
 template <typename Context>
 class MallocOutputFile : public OutputFile<Context> {
 public:
-  MallocOutputFile(Context &ctx, std::string path, i64 filesize, i64 perm)
+  MallocOutputFile(Context &ctx, std::string path, i64 filesize, int perm)
     : OutputFile<Context>(path, filesize, false), ptr(new u8[filesize]),
       perm(perm) {
     this->buf = ptr.get();
@@ -791,7 +784,15 @@ class MallocOutputFile : public OutputFile<Context> {
 
 private:
   std::unique_ptr<u8[]> ptr;
-  i64 perm;
+  int perm;
+};
+
+template <typename Context>
+class LockingOutputFile : public OutputFile<Context> {
+public:
+  LockingOutputFile(Context &ctx, std::string path, int perm);
+  void resize(Context &ctx, i64 filesize);
+  void close(Context &ctx) override;
 };
 
 //
@@ -899,6 +900,13 @@ std::optional<std::string_view> demangle_rust(std::string_view name);
 void acquire_global_lock();
 void release_global_lock();
 
+//
+// crc32.cc
+//
+
+u32 compute_crc32(u32 crc, u8 *buf, i64 len);
+std::vector<u8> crc32_solve(u32 current, u32 desired);
+
 //
 // compress.cc
 //
@@ -963,6 +971,8 @@ class MappedFile {
 public:
   ~MappedFile() { unmap(); }
   void unmap();
+  void close_fd();
+  void reopen_fd(const std::string &path);
 
   template <typename Context>
   MappedFile *slice(Context &ctx, std::string name, u64 start, u64 size) {

common/crc32.cc

@@ -0,0 +1,60 @@
+#include "common.h"
+
+#include <tbb/parallel_for_each.h>
+#include <zlib.h>
+
+namespace mold {
+
+// This function "forges" a CRC. That is, given the current and a desired
+// CRC32 value, crc32_solve() returns a binary blob to add to the end of
+// the original data to yield the desired CRC. Trailing garbage is ignored
+// by many bianry file formats, so you can create a file with a desired
+// CRC using crc32_solve(). We need it for --separate-debug-file.
+std::vector<u8> crc32_solve(u32 current, u32 desired) {
+  constexpr u32 poly = 0xedb88320;
+  u32 x = ~desired;
+
+  // Each iteration computes x = (x * x^-1) mod poly.
+  for (i64 i = 0; i < 32; i++) {
+    x = std::rotl(x, 1);
+    x ^= (x & 1) * (poly << 1);
+  }
+
+  x ^= ~current;
+
+  std::vector<u8> out(4);
+  out[0] = x;
+  out[1] = x >> 8;
+  out[2] = x >> 16;
+  out[3] = x >> 24;
+  return out;
+}
+
+// Compute a CRC for given data in parallel
+u32 compute_crc32(u32 crc, u8 *buf, i64 len) {
+  struct Shard {
+    u8 *buf;
+    i64 len;
+    u32 crc;
+  };
+
+  constexpr i64 shard_size = 1024 * 1024; // 1 MiB
+  std::vector<Shard> shards;
+
+  while (len > 0) {
+    i64 sz = std::min(len, shard_size);
+    shards.push_back({buf, sz, 0});
+    buf += sz;
+    len -= sz;
+  }
+
+  tbb::parallel_for_each(shards.begin(), shards.end(), [](Shard &shard) {
+    shard.crc = crc32_z(0, shard.buf, shard.len);
+  });
+
+  for (Shard &shard : shards)
+    crc = crc32_combine(crc, shard.crc, shard.len);
+  return crc;
+}
+
+} // namespace mold

common/filepath.cc

@@ -36,14 +36,9 @@ std::filesystem::path to_abs_path(std::filesystem::path path) {
 
 // Returns the path of the mold executable itself
 std::string get_self_path() {
-#if __APPLE__
-    char path[8192];
-    u32 size = sizeof(path);
-    if (_NSGetExecutablePath(path, &size)) {
-      std::cerr << "_NSGetExecutablePath failed\n";
-      exit(1);
-    }
-    return path;
+#if __APPLE__ || _WIN32
+  fprintf(stderr, "mold: get_self_path is not supported");
+  exit(1);
 #elif __FreeBSD__
   // /proc may not be mounted on FreeBSD. The proper way to get the
   // current executable's path is to use sysctl(2).