Additional fixes

Author: StanFromIreland

.github/workflows/reusable-san.yml

@@ -75,12 +75,11 @@ jobs:
         echo "MSAN_OPTIONS=${SAN_LOG_OPTION} allocator_may_return_null=1" >> "$GITHUB_ENV"
         # MSan reports false positives for memory initialized by libraries
         # that are not built with MSan, so disable modules that use them.
-        # _remote_debugging is disabled because it links to libzstd directly.
+        # _remote_debugging links to libzstd directly, but we unpoision the memory.
         {
           echo '*disabled*'
           echo '_bz2 _ctypes _curses _curses_panel _dbm _decimal _gdbm _hashlib'
-          echo '_lzma _remote_debugging _sqlite3 _ssl _tkinter _uuid _zstd'
-          echo 'readline zlib'
+          echo '_lzma _sqlite3 _ssl _tkinter _uuid _zstd readline zlib'
         } > Modules/Setup.local
       env:
         SAN_LOG_OPTION: log_path=${{ github.workspace }}/san_log
@@ -108,6 +107,8 @@ jobs:
     # gh-157958: -O2 instead of the pydebug default -Og to avoid a clang 21
     # compile-time blowup on some interpreter files.
     # (https://github.com/llvm/llvm-project/issues/179695)
+    # MSan uses --with-assertions instead of --with-pydebug because its
+    # hooks on the Python memory allocators hide uninitialized reads.
     - name: Configure CPython
       run: >-
         ./configure
@@ -120,7 +121,7 @@ jobs:
           && '--with-memory-sanitizer'
           || '--with-undefined-behavior-sanitizer --with-strict-overflow'
         }}
-        --with-pydebug
+        ${{ inputs.sanitizer == 'MSan' && '--with-assertions' || '--with-pydebug' }}
         ${{ inputs.sanitizer == 'TSan' && '--with-openssl="$OPENSSL_DIR" --with-openssl-rpath=auto' || '' }}
         ${{ inputs.free-threading && '--disable-gil' || '' }}
     - name: Build CPython

Lib/test/test_asyncio/test_tools.py

@@ -1,11 +1,9 @@
 import unittest
 
-from test.support import import_helper
-
-_remote_debugging = import_helper.import_module('_remote_debugging')
-
 from asyncio import tools
 
+import _remote_debugging
+
 
 def LocationInfo(lineno, end_lineno=None, col_offset=None, end_col_offset=None):
     return _remote_debugging.LocationInfo((lineno, end_lineno, col_offset, end_col_offset))

Modules/_remote_debugging/binary_io_reader.c

@@ -19,6 +19,10 @@
 #include <zstd.h>
 #endif
 
+#ifdef _Py_MEMORY_SANITIZER
+#  include <sanitizer/msan_interface.h>
+#endif
+
 /* ============================================================================
  * CONSTANTS FOR BINARY FORMAT SIZES
  * ============================================================================ */
@@ -315,6 +319,7 @@ reader_decompress_samples(BinaryReader *reader, const uint8_t *data)
             return -1;
         }
 
+        _Py_MSAN_UNPOISON(output.dst, output.pos);
         total_output += output.pos;
     }
 

Modules/_remote_debugging/binary_io_writer.c

@@ -19,6 +19,10 @@
 #include <zstd.h>
 #endif
 
+#ifdef _Py_MEMORY_SANITIZER
+#  include <sanitizer/msan_interface.h>
+#endif
+
 /* ============================================================================
  * CONSTANTS FOR BINARY FORMAT SIZES
  * ============================================================================ */
@@ -235,6 +239,7 @@ writer_flush_buffer(BinaryWriter *writer)
                 return -1;
             }
 
+            _Py_MSAN_UNPOISON(writer->zstd.compressed_buffer, output.pos);
             if (output.pos > 0) {
                 if (fwrite_checked_allow_threads(writer->zstd.compressed_buffer, output.pos, writer->fp) < 0) {
                     return -1;
@@ -1084,6 +1089,7 @@ binary_writer_finalize(BinaryWriter *writer)
                 return -1;
             }
 
+            _Py_MSAN_UNPOISON(writer->zstd.compressed_buffer, output.pos);
             if (output.pos > 0) {
                 if (fwrite_checked_allow_threads(writer->zstd.compressed_buffer, output.pos, writer->fp) < 0) {
                     return -1;

Python/instrumentation.c

@@ -1690,6 +1690,7 @@ allocate_instrumentation_data(PyCodeObject *code)
         }
         monitoring->local_monitors = (_Py_LocalMonitors){ 0 };
         monitoring->active_monitors = (_Py_LocalMonitors){ 0 };
+        memset(monitoring->tool_versions, 0, sizeof(monitoring->tool_versions));
         monitoring->tools = NULL;
         monitoring->lines = NULL;
         monitoring->line_tools = NULL;