Merge branch 'main' into fix/sysconfig/relic-128978

Author: picnixz

.github/workflows/build.yml

@@ -231,10 +231,14 @@ jobs:
     name: >-
       Ubuntu
       ${{ fromJSON(matrix.free-threading) && '(free-threading)' || '' }}
+      ${{ fromJSON(matrix.bolt) && '(bolt)' || '' }}
     needs: check_source
     if: needs.check_source.outputs.run_tests == 'true'
     strategy:
       matrix:
+        bolt:
+        - false
+        - true
         free-threading:
         - false
         - true
@@ -246,9 +250,16 @@ jobs:
         exclude:
         - os: ubuntu-24.04-aarch64
           is-fork: true
+        # Do not test BOLT with free-threading, to conserve resources
+        - bolt: true
+          free-threading: true
+        # BOLT currently crashes during instrumentation on aarch64
+        - os: ubuntu-24.04-aarch64
+          bolt: true
     uses: ./.github/workflows/reusable-ubuntu.yml
     with:
       config_hash: ${{ needs.check_source.outputs.config_hash }}
+      bolt-optimizations: ${{ matrix.bolt }}
       free-threading: ${{ matrix.free-threading }}
       os: ${{ matrix.os }}
 

.github/workflows/reusable-ubuntu.yml

@@ -6,6 +6,11 @@ on:
       config_hash:
         required: true
         type: string
+      bolt-optimizations:
+        description: Whether to enable BOLT optimizations
+        required: false
+        type: boolean
+        default: false
       free-threading:
         description: Whether to use free-threaded mode
         required: false
@@ -34,6 +39,12 @@ jobs:
       run: echo "::add-matcher::.github/problem-matchers/gcc.json"
     - name: Install dependencies
       run: sudo ./.github/workflows/posix-deps-apt.sh
+    - name: Install Clang and BOLT
+      if: ${{ fromJSON(inputs.bolt-optimizations) }}
+      run: |
+        sudo bash -c "$(wget -O - https://apt.llvm.org/llvm.sh)" ./llvm.sh 19
+        sudo apt-get install bolt-19
+        echo PATH="$(llvm-config-19 --bindir):$PATH" >> $GITHUB_ENV
     - name: Configure OpenSSL env vars
       run: |
         echo "MULTISSL_DIR=${GITHUB_WORKSPACE}/multissl" >> "$GITHUB_ENV"
@@ -73,14 +84,18 @@ jobs:
         key: ${{ github.job }}-${{ runner.os }}-${{ env.IMAGE_VERSION }}-${{ inputs.config_hash }}
     - name: Configure CPython out-of-tree
       working-directory: ${{ env.CPYTHON_BUILDDIR }}
+      # `test_unpickle_module_race` writes to the source directory, which is
+      # read-only during builds — so we exclude it from profiling with BOLT.
       run: >-
+        PROFILE_TASK='-m test --pgo --ignore test_unpickle_module_race'
         ../cpython-ro-srcdir/configure
         --config-cache
         --with-pydebug
         --enable-slower-safety
         --enable-safety
         --with-openssl="$OPENSSL_DIR"
         ${{ fromJSON(inputs.free-threading) && '--disable-gil' || '' }}
+        ${{ fromJSON(inputs.bolt-optimizations) && '--enable-bolt' || '' }}
     - name: Build CPython out-of-tree
       if: ${{ inputs.free-threading }}
       working-directory: ${{ env.CPYTHON_BUILDDIR }}

Doc/glossary.rst

@@ -115,7 +115,7 @@ Glossary
       :keyword:`yield` expression.
 
       Each :keyword:`yield` temporarily suspends processing, remembering the
-      location execution state (including local variables and pending
+      execution state (including local variables and pending
       try-statements).  When the *asynchronous generator iterator* effectively
       resumes with another awaitable returned by :meth:`~object.__anext__`, it
       picks up where it left off.  See :pep:`492` and :pep:`525`.
@@ -564,7 +564,7 @@ Glossary
       An object created by a :term:`generator` function.
 
       Each :keyword:`yield` temporarily suspends processing, remembering the
-      location execution state (including local variables and pending
+      execution state (including local variables and pending
       try-statements).  When the *generator iterator* resumes, it picks up where
       it left off (in contrast to functions which start fresh on every
       invocation).

Doc/library/importlib.rst

@@ -380,13 +380,15 @@ ABC hierarchy::
 
 .. class:: ResourceLoader
 
+   *Superseded by TraversableResources*
+
     An abstract base class for a :term:`loader` which implements the optional
     :pep:`302` protocol for loading arbitrary resources from the storage
     back-end.
 
     .. deprecated:: 3.7
        This ABC is deprecated in favour of supporting resource loading
-       through :class:`importlib.resources.abc.ResourceReader`.
+       through :class:`importlib.resources.abc.TraversableResources`.
 
     .. abstractmethod:: get_data(path)
 

Doc/library/sys.rst

@@ -8,7 +8,7 @@
 
 This module provides access to some variables used or maintained by the
 interpreter and to functions that interact strongly with the interpreter. It is
-always available.
+always available. Unless explicitly noted otherwise, all variables are read-only.
 
 
 .. data:: abiflags

Include/cpython/object.h

@@ -475,9 +475,6 @@ partially-deallocated object. To check this, the tp_dealloc function must be
 passed as second argument to Py_TRASHCAN_BEGIN().
 */
 
-/* Python 3.9 private API, invoked by the macros below. */
-PyAPI_FUNC(int) _PyTrash_begin(PyThreadState *tstate, PyObject *op);
-PyAPI_FUNC(void) _PyTrash_end(PyThreadState *tstate);
 
 PyAPI_FUNC(void) _PyTrash_thread_deposit_object(PyThreadState *tstate, PyObject *op);
 PyAPI_FUNC(void) _PyTrash_thread_destroy_chain(PyThreadState *tstate);

Include/internal/pycore_code.h

@@ -100,6 +100,7 @@ typedef struct {
 
 typedef struct {
     _Py_BackoffCounter counter;
+    uint16_t external_cache[4];
 } _PyBinaryOpCache;
 
 #define INLINE_CACHE_ENTRIES_BINARY_OP CACHE_ENTRIES(_PyBinaryOpCache)
@@ -438,7 +439,7 @@ write_u64(uint16_t *p, uint64_t val)
 }
 
 static inline void
-write_obj(uint16_t *p, PyObject *val)
+write_ptr(uint16_t *p, void *val)
 {
     memcpy(p, &val, sizeof(val));
 }
@@ -576,6 +577,16 @@ adaptive_counter_backoff(_Py_BackoffCounter counter) {
     return restart_backoff_counter(counter);
 }
 
+/* Specialization Extensions */
+
+/* callbacks for an external specialization */
+typedef int (*binaryopguardfunc)(PyObject *lhs, PyObject *rhs);
+typedef PyObject *(*binaryopactionfunc)(PyObject *lhs, PyObject *rhs);
+
+typedef struct {
+    binaryopguardfunc guard;
+    binaryopactionfunc action;
+} _PyBinaryOpSpecializationDescr;
 
 /* Comparison bit masks. */
 

Include/internal/pycore_dict.h

@@ -347,8 +347,7 @@ PyDictObject *_PyObject_MaterializeManagedDict_LockHeld(PyObject *);
 static inline Py_ssize_t
 _PyDict_UniqueId(PyDictObject *mp)
 {
-    // Offset by one so that _ma_watcher_tag=0 represents an unassigned id
-    return (Py_ssize_t)(mp->_ma_watcher_tag >> DICT_UNIQUE_ID_SHIFT) - 1;
+    return (Py_ssize_t)(mp->_ma_watcher_tag >> DICT_UNIQUE_ID_SHIFT);
 }
 
 static inline void

Include/internal/pycore_gc.h

@@ -45,12 +45,13 @@ static inline PyObject* _Py_FROM_GC(PyGC_Head *gc) {
  * the per-object lock.
  */
 #ifdef Py_GIL_DISABLED
-#  define _PyGC_BITS_TRACKED        (1)     // Tracked by the GC
-#  define _PyGC_BITS_FINALIZED      (2)     // tp_finalize was called
-#  define _PyGC_BITS_UNREACHABLE    (4)
-#  define _PyGC_BITS_FROZEN         (8)
-#  define _PyGC_BITS_SHARED         (16)
-#  define _PyGC_BITS_DEFERRED       (64)    // Use deferred reference counting
+#  define _PyGC_BITS_TRACKED        (1<<0)     // Tracked by the GC
+#  define _PyGC_BITS_FINALIZED      (1<<1)     // tp_finalize was called
+#  define _PyGC_BITS_UNREACHABLE    (1<<2)
+#  define _PyGC_BITS_FROZEN         (1<<3)
+#  define _PyGC_BITS_SHARED         (1<<4)
+#  define _PyGC_BITS_ALIVE          (1<<5)    // Reachable from a known root.
+#  define _PyGC_BITS_DEFERRED       (1<<6)    // Use deferred reference counting
 #endif
 
 #ifdef Py_GIL_DISABLED
@@ -330,6 +331,9 @@ struct _gc_runtime_state {
        collections, and are awaiting to undergo a full collection for
        the first time. */
     Py_ssize_t long_lived_pending;
+
+    /* True if gc.freeze() has been used. */
+    int freeze_active;
 #endif
 };
 

Include/internal/pycore_magic_number.h

@@ -266,6 +266,7 @@ Known values:
     Python 3.14a4 3611 (Add NOT_TAKEN instruction)
     Python 3.14a4 3612 (Add POP_ITER and INSTRUMENTED_POP_ITER)
     Python 3.14a4 3613 (Add LOAD_CONST_MORTAL instruction)
+    Python 3.14a5 3614 (Add BINARY_OP_EXTEND)
 
     Python 3.15 will start with 3650
 
@@ -278,7 +279,7 @@ PC/launcher.c must also be updated.
 
 */
 
-#define PYC_MAGIC_NUMBER 3613
+#define PYC_MAGIC_NUMBER 3614
 /* This is equivalent to converting PYC_MAGIC_NUMBER to 2 bytes
    (little-endian) and then appending b'\r\n'. */
 #define PYC_MAGIC_NUMBER_TOKEN \