[pull] main from python:main

.github/workflows/build.yml

@@ -165,13 +165,21 @@ jobs:
         free-threading:
           - false
           - true
+        interpreter:
+          - switch-case
         exclude:
           # Skip Win32 on free-threaded builds
           - { arch: Win32, free-threading: true }
+        include:
+          # msvc::musttail is currently only supported on x64,
+          # and only supported on 3.15+.
+          - { arch: x64, free-threading: false, interpreter: tail-call }
+          - { arch: x64, free-threading: true,  interpreter: tail-call }
     uses: ./.github/workflows/reusable-windows.yml
     with:
       arch: ${{ matrix.arch }}
       free-threading: ${{ matrix.free-threading }}
+      interpreter: ${{ matrix.interpreter }}
 
   build-windows-msi:
     # ${{ '' } is a hack to nest jobs under the same sidebar category.

.github/workflows/reusable-windows.yml

@@ -12,6 +12,10 @@ on:
         required: false
         type: boolean
         default: false
+      interpreter:
+        description: Which interpreter to build (switch-case or tail-call)
+        required: true
+        type: string
 
 env:
   FORCE_COLOR: 1
@@ -20,7 +24,7 @@ env:
 
 jobs:
   build:
-    name: Build and test (${{ inputs.arch }})
+    name: Build and test (${{ inputs.arch }}, ${{ inputs.interpreter }})
     runs-on: ${{ inputs.arch == 'arm64' && 'windows-11-arm' || 'windows-2025-vs2026' }}
     timeout-minutes: 60
     env:
@@ -33,9 +37,12 @@ jobs:
       if: inputs.arch != 'Win32'
       run: echo "::add-matcher::.github/problem-matchers/msvc.json"
     - name: Build CPython
+      # msvc::musttail is not supported for debug builds, so we have to
+      # switch to release.
       run: >-
         .\\PCbuild\\build.bat
-        -e -d -v
+        -e -v
+        ${{ inputs.interpreter == 'switch-case' && '-d' || '--tail-call-interp -c Release' }}
         -p "${ARCH}"
         ${{ fromJSON(inputs.free-threading) && '--disable-gil' || '' }}
       shell: bash
@@ -45,6 +52,7 @@ jobs:
       run: >-
         .\\PCbuild\\rt.bat
         -p "${ARCH}"
-        -d -q --fast-ci
+        -q --fast-ci
+        ${{ inputs.interpreter == 'switch-case' && '-d' || '' }}
         ${{ fromJSON(inputs.free-threading) && '--disable-gil' || '' }}
       shell: bash

.github/workflows/tail-call.yml

@@ -23,41 +23,6 @@ env:
   LLVM_VERSION: 21
 
 jobs:
-  windows:
-    name: ${{ matrix.target }}
-    runs-on: ${{ matrix.runner }}
-    timeout-minutes: 60
-    strategy:
-      fail-fast: false
-      matrix:
-        include:
-          - target: x86_64-pc-windows-msvc/msvc
-            architecture: x64
-            runner: windows-2025-vs2026
-            build_flags: ""
-            run_tests: true
-          - target: x86_64-pc-windows-msvc/msvc-free-threading
-            architecture: x64
-            runner: windows-2025-vs2026
-            build_flags: --disable-gil
-            run_tests: false
-    steps:
-      - uses: actions/checkout@v6
-        with:
-          persist-credentials: false
-      - uses: actions/setup-python@v6
-        with:
-          python-version: '3.11'
-      - name: Build
-        shell: pwsh
-        run: |
-          ./PCbuild/build.bat --tail-call-interp ${{ matrix.build_flags }} -c Release -p ${{ matrix.architecture }}
-      - name: Test
-        if: matrix.run_tests
-        shell: pwsh
-        run: |
-          ./PCbuild/rt.bat -p ${{ matrix.architecture }} -q --multiprocess 0 --timeout 4500 --verbose2 --verbose3
-
   macos:
     name: ${{ matrix.target }}
     runs-on: ${{ matrix.runner }}

Doc/library/profiling.sampling.rst

@@ -1003,6 +1003,47 @@ at the top indicate functions that consume significant time either directly
 or through their callees.
 
 
+Differential flame graphs
+~~~~~~~~~~~~~~~~~~~~~~~~~
+
+Differential flame graphs compare two profiling runs to highlight where
+performance changed. This helps identify regressions introduced by code changes
+and validate that optimizations achieved their intended effect::
+
+   # Capture baseline profile
+   python -m profiling.sampling run --binary -o baseline.bin script.py
+
+   # After modifying code, generate differential flamegraph
+   python -m profiling.sampling run --diff-flamegraph baseline.bin -o diff.html script.py
+
+The visualization draws the current profile with frame widths showing current
+time consumption, then applies color to indicate how each function changed
+relative to the baseline.
+
+**Color coding**:
+
+- **Red**: Functions consuming more time (regressions). Lighter shades indicate
+  modest increases, while darker shades show severe regressions.
+
+- **Blue**: Functions consuming less time (improvements). Lighter shades for
+  modest reductions, darker shades for significant speedups.
+
+- **Gray**: Minimal or no change.
+
+- **Purple**: New functions not present in the baseline.
+
+Frame colors indicate changes in **direct time** (time when the function was at
+the top of the stack, actively executing), not cumulative time including callees.
+Hovering over a frame shows comparison details including baseline time, current
+time, and the percentage change.
+
+Some call paths may disappear entirely between profiles. These are called
+**elided stacks** and occur when optimizations eliminate code paths or certain
+branches stop executing. If elided stacks are present, an elided toggle appears
+allowing you to switch between the main differential view and an elided-only
+view that shows just the removed paths (colored purple).
+
+
 Gecko format
 ------------
 
@@ -1488,6 +1529,10 @@ Output options
 
    Generate self-contained HTML flame graph.
 
+.. option:: --diff-flamegraph <baseline.bin>
+
+   Generate differential flamegraph comparing to a baseline binary profile.
+
 .. option:: --gecko
 
    Generate Gecko JSON format for Firefox Profiler.

Doc/library/xml.etree.elementtree.rst

@@ -691,7 +691,7 @@ Functions
    .. versionadded:: 3.2
 
 
-.. function:: SubElement(parent, tag, attrib={}, **extra)
+.. function:: SubElement(parent, tag, /, attrib={}, **extra)
 
    Subelement factory.  This function creates an element instance, and appends
    it to an existing element.
@@ -705,6 +705,9 @@ Functions
    .. versionchanged:: 3.15
       *attrib* can now be a :class:`frozendict`.
 
+   .. versionchanged:: next
+      *parent* and *tag* are now positional-only parameters.
+
 
 .. function:: tostring(element, encoding="us-ascii", method="xml", *, \
                        xml_declaration=None, default_namespace=None, \
@@ -880,7 +883,7 @@ Element Objects
    :noindex:
    :no-index:
 
-.. class:: Element(tag, attrib={}, **extra)
+.. class:: Element(tag, /, attrib={}, **extra)
 
    Element class.  This class defines the Element interface, and provides a
    reference implementation of this interface.
@@ -893,6 +896,9 @@ Element Objects
    .. versionchanged:: 3.15
       *attrib* can now be a :class:`frozendict`.
 
+   .. versionchanged:: next
+      *tag* is now a positional-only parameter.
+
 
    .. attribute:: tag
 

Include/internal/pycore_dict.h

@@ -138,13 +138,14 @@ extern PyObject *_PyDict_LoadBuiltinsFromGlobals(PyObject *globals);
 
 /* Consumes references to key and value */
 PyAPI_FUNC(int) _PyDict_SetItem_Take2(PyDictObject *op, PyObject *key, PyObject *value);
+PyAPI_FUNC(int) _PyDict_SetItem_Take2_KnownHash(PyDictObject *op, PyObject *key, PyObject *value, Py_hash_t hash);
 extern int _PyDict_SetItem_LockHeld(PyDictObject *dict, PyObject *name, PyObject *value);
 // Export for '_asyncio' shared extension
 PyAPI_FUNC(int) _PyDict_SetItem_KnownHash_LockHeld(PyDictObject *mp, PyObject *key,
                                                    PyObject *value, Py_hash_t hash);
 // Export for '_asyncio' shared extension
 PyAPI_FUNC(int) _PyDict_GetItemRef_KnownHash_LockHeld(PyDictObject *op, PyObject *key, Py_hash_t hash, PyObject **result);
-extern int _PyDict_GetItemRef_KnownHash(PyDictObject *op, PyObject *key, Py_hash_t hash, PyObject **result);
+PyAPI_FUNC(int) _PyDict_GetItemRef_KnownHash(PyDictObject *op, PyObject *key, Py_hash_t hash, PyObject **result);
 extern int _PyDict_GetItemRef_Unicode_LockHeld(PyDictObject *op, PyObject *key, PyObject **result);
 PyAPI_FUNC(int) _PyObjectDict_SetItem(PyTypeObject *tp, PyObject *obj, PyObject **dictptr, PyObject *name, PyObject *value);
 

Include/internal/pycore_long.h

@@ -64,7 +64,8 @@ PyAPI_FUNC(void) _PyLong_ExactDealloc(PyObject *self);
 #  error "_PY_NSMALLPOSINTS must be greater than or equal to 257"
 #endif
 
-#define _PY_IS_SMALL_INT(val) ((val) >= 0 && (val) < 256 && (val) < _PY_NSMALLPOSINTS)
+#define _PY_IS_SMALL_INT(val) \
+    (-_PY_NSMALLNEGINTS <= (val) && (val) < _PY_NSMALLPOSINTS)
 
 // Return a reference to the immortal zero singleton.
 // The function cannot return NULL.