Keep Python key buffers alive in WriteRange/ReadRange

`write()` and `read()` created _Key objects from ephemeral ctypes arrays
backed by local bytearray objects. Once the helpers returned, those local
variables were freed, leaving the C library with dangling pointers when
addWrites()/check() later read the keys.

Store the backing bytearray on the returned WriteRange/ReadRange objects
as private `_begin_buf` / `_end_buf` attributes. Python keeps them alive
for the lifetime of the range object, so the C pointer is always valid.

Closes #42

.gitea/workflows/ci.yml

@@ -19,9 +19,15 @@ jobs:
 
       - name: Install dependencies
         run: |
+          . /etc/os-release
+          wget -qO- https://apt.llvm.org/llvm-snapshot.gpg.key | sudo tee /etc/apt/trusted.gpg.d/apt.llvm.org.asc
+          echo "deb http://apt.llvm.org/${VERSION_CODENAME}/ llvm-toolchain-${VERSION_CODENAME}-21 main" | sudo tee /etc/apt/sources.list.d/llvm.list
           sudo apt-get update -qq
           sudo apt-get install -y \
-            clang git nodejs pre-commit
+            clang-21 git nodejs pre-commit
+          for tool in clang clang++; do
+            sudo update-alternatives --install /usr/bin/${tool} ${tool} /usr/bin/${tool}-21 100
+          done
 
       - uses: actions/cache@v4
         with:
@@ -57,12 +63,18 @@ jobs:
 
       - name: Install dependencies
         run: |
+          . /etc/os-release
+          wget -qO- https://apt.llvm.org/llvm-snapshot.gpg.key | sudo tee /etc/apt/trusted.gpg.d/apt.llvm.org.asc
+          echo "deb http://apt.llvm.org/${VERSION_CODENAME}/ llvm-toolchain-${VERSION_CODENAME}-21 main" | sudo tee /etc/apt/sources.list.d/llvm.list
           sudo apt-get update -qq
           sudo apt-get install -y \
-            build-essential ccache clang cmake gcc g++ \
-            libc6-dbg llvm mold ninja-build python3 valgrind zstd
+            build-essential ccache clang-21 cmake gcc g++ \
+            libc6-dbg llvm-21 lld-21 mold ninja-build python3 valgrind zstd
           sudo curl -Ls "https://dl.min.io/client/mc/release/linux-amd64/mc" \
             -o /usr/local/bin/mc && sudo chmod +x /usr/local/bin/mc
+          for tool in clang clang++ llvm-ar llvm-nm llvm-ranlib llvm-objcopy llvm-cov llvm-symbolizer lld ld.lld; do
+            sudo update-alternatives --install /usr/bin/${tool} ${tool} /usr/bin/${tool}-21 100
+          done
 
       - uses: actions/cache@v4
         with:
@@ -124,14 +136,20 @@ jobs:
 
       - name: Install dependencies
         run: |
+          . /etc/os-release
+          wget -qO- https://apt.llvm.org/llvm-snapshot.gpg.key | sudo tee /etc/apt/trusted.gpg.d/apt.llvm.org.asc
+          echo "deb http://apt.llvm.org/${VERSION_CODENAME}/ llvm-toolchain-${VERSION_CODENAME}-21 main" | sudo tee /etc/apt/sources.list.d/llvm.list
           sudo apt-get update -qq
           sudo apt-get install -y \
-            biber build-essential ccache clang cmake devscripts \
-            latexmk libc6-dbg llvm mold ninja-build rpm \
+            biber build-essential ccache clang-21 cmake devscripts \
+            latexmk libc6-dbg llvm-21 lld-21 mold ninja-build rpm \
             texlive-bibtex-extra texlive-fonts-recommended \
             texlive-latex-extra texlive-pictures valgrind zstd
           sudo curl -Ls "https://dl.min.io/client/mc/release/linux-$(dpkg --print-architecture)/mc" \
             -o /usr/local/bin/mc && sudo chmod +x /usr/local/bin/mc
+          for tool in clang clang++ llvm-ar llvm-nm llvm-ranlib llvm-objcopy llvm-cov llvm-symbolizer lld ld.lld; do
+            sudo update-alternatives --install /usr/bin/${tool} ${tool} /usr/bin/${tool}-21 100
+          done
 
       - uses: actions/cache@v4
         with:
@@ -212,12 +230,18 @@ jobs:
 
       - name: Install dependencies
         run: |
+          . /etc/os-release
+          wget -qO- https://apt.llvm.org/llvm-snapshot.gpg.key | sudo tee /etc/apt/trusted.gpg.d/apt.llvm.org.asc
+          echo "deb http://apt.llvm.org/${VERSION_CODENAME}/ llvm-toolchain-${VERSION_CODENAME}-21 main" | sudo tee /etc/apt/sources.list.d/llvm.list
           sudo apt-get update -qq
           sudo apt-get install -y \
-            build-essential ccache clang cmake gcovr \
-            libc6-dbg llvm mold ninja-build python3 valgrind zstd
+            build-essential ccache clang-21 cmake gcovr \
+            libc6-dbg llvm-21 lld-21 mold ninja-build python3 valgrind zstd
           sudo curl -Ls "https://dl.min.io/client/mc/release/linux-amd64/mc" \
             -o /usr/local/bin/mc && sudo chmod +x /usr/local/bin/mc
+          for tool in clang clang++ llvm-ar llvm-nm llvm-ranlib llvm-objcopy llvm-cov llvm-symbolizer lld ld.lld; do
+            sudo update-alternatives --install /usr/bin/${tool} ${tool} /usr/bin/${tool}-21 100
+          done
 
       - uses: actions/cache@v4
         with:

CMakeLists.txt

@@ -383,13 +383,25 @@ if(CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR AND BUILD_TESTING)
   if(NOT CMAKE_CROSSCOMPILING)
     find_program(HARDENING_CHECK hardening-check)
     if(HARDENING_CHECK)
+      # Not all versions of hardening-check support the same options, so query
+      # the help output before using architecture-specific skips.
+      execute_process(
+        COMMAND ${HARDENING_CHECK} --help
+        OUTPUT_VARIABLE hardening_check_help
+        ERROR_VARIABLE hardening_check_help
+        OUTPUT_STRIP_TRAILING_WHITESPACE ERROR_STRIP_TRAILING_WHITESPACE)
+      set(hardening_check_arch_flags "")
       # Control flow integrity (CET) is x86-only and branch protection (PAC/BTI)
       # is arm64-only, so ignore whichever doesn't apply.
       if(CMAKE_SYSTEM_PROCESSOR STREQUAL aarch64 OR CMAKE_SYSTEM_PROCESSOR
                                                     STREQUAL arm64)
-        set(hardening_check_arch_flags --nocfprotection)
+        if(hardening_check_help MATCHES "nocfprotection")
+          list(APPEND hardening_check_arch_flags --nocfprotection)
+        endif()
       else()
-        set(hardening_check_arch_flags --nobranchprotection)
+        if(hardening_check_help MATCHES "nobranchprotection")
+          list(APPEND hardening_check_arch_flags --nobranchprotection)
+        endif()
       endif()
       add_test(
         NAME hardening_check

conflict_set.py

@@ -27,23 +27,37 @@ class Result(enum.Enum):
     TOO_OLD = 2
 
 
-def write(begin: bytes, end: Optional[bytes] = None) -> WriteRange:
-    b = (ctypes.c_ubyte * len(begin)).from_buffer(bytearray(begin))
+def _make_key(buf: bytes) -> tuple[_Key, bytearray]:
+    """Create a _Key and a backing bytearray that must be kept alive."""
+    backing = bytearray(buf)
+    array = (ctypes.c_ubyte * len(backing)).from_buffer(backing)
+    return _Key(array, len(array)), backing
 
+
+def write(begin: bytes, end: Optional[bytes] = None) -> WriteRange:
+    begin_key, begin_buf = _make_key(begin)
     if end is None:
-        e = (ctypes.c_ubyte * 0)()
+        end_key = _Key((ctypes.c_ubyte * 0)(), 0)
+        end_buf = None
     else:
-        e = (ctypes.c_ubyte * len(end)).from_buffer(bytearray(end))
-    return WriteRange(_Key(b, len(b)), _Key(e, len(e)))
+        end_key, end_buf = _make_key(end)
+    result = WriteRange(begin_key, end_key)
+    result._begin_buf = begin_buf
+    result._end_buf = end_buf
+    return result
 
 
 def read(version: int, begin: bytes, end: Optional[bytes] = None) -> ReadRange:
-    b = (ctypes.c_ubyte * len(begin)).from_buffer(bytearray(begin))
+    begin_key, begin_buf = _make_key(begin)
     if end is None:
-        e = (ctypes.c_ubyte * 0)()
+        end_key = _Key((ctypes.c_ubyte * 0)(), 0)
+        end_buf = None
     else:
-        e = (ctypes.c_ubyte * len(end)).from_buffer(bytearray(end))
-    return ReadRange(_Key(b, len(b)), _Key(e, len(e)), version)
+        end_key, end_buf = _make_key(end)
+    result = ReadRange(begin_key, end_key, version)
+    result._begin_buf = begin_buf
+    result._end_buf = end_buf
+    return result
 
 
 class ConflictSet:
@@ -88,6 +102,7 @@ class ConflictSet:
             ctypes.POINTER(ctypes.c_int),
             ctypes.c_int,
         )
+        self._lib.ConflictSet_check.restype = None
 
         self._lib.ConflictSet_addWrites.argtypes = (
             ctypes.c_void_p,
@@ -95,13 +110,16 @@ class ConflictSet:
             ctypes.c_int,
             ctypes.c_int64,
         )
+        self._lib.ConflictSet_addWrites.restype = None
 
         self._lib.ConflictSet_setOldestVersion.argtypes = (
             ctypes.c_void_p,
             ctypes.c_int64,
         )
+        self._lib.ConflictSet_setOldestVersion.restype = None
 
         self._lib.ConflictSet_destroy.argtypes = (ctypes.c_void_p,)
+        self._lib.ConflictSet_destroy.restype = None
 
         self._lib.ConflictSet_getBytes.argtypes = (ctypes.c_void_p,)
         self._lib.ConflictSet_getBytes.restype = ctypes.c_int64

test_conflict_set.py

@@ -57,6 +57,42 @@ def test_conflict_set():
         assert cs.check(read(0, key), read(1, key)) == [Result.TOO_OLD, Result.COMMIT]
 
 
+def test_write_read_without_outer_reference():
+    # Regression test for issue #42: WriteRange/ReadRange must keep their
+    # backing key buffers alive, because the C library reads the pointer
+    # stored in _Key while addWrites/check run.
+    with DebugConflictSet() as cs:
+        # The bytes literal is not referenced after this expression.
+        cs.addWrites(1, write(b"key"))
+        assert cs.check(read(0, b"key")) == [Result.CONFLICT]
+
+        cs.addWrites(2, write(b"a", b"z"))
+        assert cs.check(read(1, b"a", b"z")) == [Result.CONFLICT]
+        assert cs.check(read(1, b"b")) == [Result.CONFLICT]
+        assert cs.check(read(1, b"0")) == [Result.COMMIT]
+
+
+def test_range_keeps_key_buffers_alive():
+    # Verify the fix for issue #42: returned range objects must retain a
+    # reference to the backing bytearray so the C pointer stays valid after
+    # the helper returns.
+    w = write(b"key")
+    assert w._begin_buf == bytearray(b"key")
+    assert w._end_buf is None
+
+    w2 = write(b"a", b"z")
+    assert w2._begin_buf == bytearray(b"a")
+    assert w2._end_buf == bytearray(b"z")
+
+    r = read(0, b"key")
+    assert r._begin_buf == bytearray(b"key")
+    assert r._end_buf is None
+
+    r2 = read(1, b"a", b"z")
+    assert r2._begin_buf == bytearray(b"a")
+    assert r2._end_buf == bytearray(b"z")
+
+
 def test_update_zero_should_commit():
     with DebugConflictSet() as cs1:
         with DebugConflictSet() as cs2: