Merge branch 'main' into cmaloney/pyio_fileio_readall

Author: cmaloney

.github/workflows/build.yml

@@ -512,26 +512,20 @@ jobs:
       run: xvfb-run make ci
 
   build_tsan:
-    name: 'Thread sanitizer'
-    needs: check_source
-    if: needs.check_source.outputs.run_tests == 'true'
-    uses: ./.github/workflows/reusable-tsan.yml
-    with:
-      config_hash: ${{ needs.check_source.outputs.config_hash }}
-      options: ./configure --config-cache --with-thread-sanitizer --with-pydebug
-      suppressions_path: Tools/tsan/supressions.txt
-      tsan_logs_artifact_name: tsan-logs-default
-
-  build_tsan_free_threading:
-    name: 'Thread sanitizer (free-threading)'
+    name: >-
+      Thread sanitizer
+      ${{ fromJSON(matrix.free-threading) && '(free-threading)' || '' }}
     needs: check_source
     if: needs.check_source.outputs.run_tests == 'true'
+    strategy:
+      matrix:
+        free-threading:
+        - false
+        - true
     uses: ./.github/workflows/reusable-tsan.yml
     with:
       config_hash: ${{ needs.check_source.outputs.config_hash }}
-      options: ./configure --config-cache --disable-gil --with-thread-sanitizer --with-pydebug
-      suppressions_path: Tools/tsan/suppressions_free_threading.txt
-      tsan_logs_artifact_name: tsan-logs-free-threading
+      free-threading: ${{ matrix.free-threading }}
 
   # CIFuzz job based on https://google.github.io/oss-fuzz/getting-started/continuous-integration/
   cifuzz:
@@ -591,7 +585,6 @@ jobs:
     - test_hypothesis
     - build_asan
     - build_tsan
-    - build_tsan_free_threading
     - cifuzz
 
     runs-on: ubuntu-latest
@@ -625,7 +618,6 @@ jobs:
             build_windows,
             build_asan,
             build_tsan,
-            build_tsan_free_threading,
             '
             || ''
           }}

.github/workflows/reusable-tsan.yml

@@ -6,17 +6,11 @@ on:
       config_hash:
         required: true
         type: string
-      options:
-        required: true
-        type: string
-      suppressions_path:
-        description: 'A repo relative path to the suppressions file'
-        required: true
-        type: string
-      tsan_logs_artifact_name:
-        description: 'Name of the TSAN logs artifact. Must be unique for each job.'
-        required: true
-        type: string
+      free-threading:
+        description: Whether to use free-threaded mode
+        required: false
+        type: boolean
+        default: false
 
 env:
   FORCE_COLOR: 1
@@ -26,9 +20,6 @@ jobs:
     name: 'Thread sanitizer'
     runs-on: ubuntu-24.04
     timeout-minutes: 60
-    env:
-      OPTIONS: ${{ inputs.options }}
-      SUPPRESSIONS_PATH: ${{ inputs.suppressions_path }}
     steps:
     - uses: actions/checkout@v4
       with:
@@ -55,7 +46,11 @@ jobs:
         sudo sysctl -w vm.mmap_rnd_bits=28
     - name: TSAN Option Setup
       run: |
-        echo "TSAN_OPTIONS=log_path=${GITHUB_WORKSPACE}/tsan_log suppressions=${GITHUB_WORKSPACE}/${SUPPRESSIONS_PATH} handle_segv=0" >> "$GITHUB_ENV"
+        echo "TSAN_OPTIONS=log_path=${GITHUB_WORKSPACE}/tsan_log suppressions=${GITHUB_WORKSPACE}/Tools/tsan/suppressions${{
+            fromJSON(inputs.free-threading)
+            && '_free_threading'
+            || ''
+          }}.txt handle_segv=0" >> "$GITHUB_ENV"
         echo "CC=clang" >> "$GITHUB_ENV"
         echo "CXX=clang++" >> "$GITHUB_ENV"
     - name: Add ccache to PATH
@@ -67,7 +62,12 @@ jobs:
         save: ${{ github.event_name == 'push' }}
         max-size: "200M"
     - name: Configure CPython
-      run: "${OPTIONS}"
+      run: >-
+        ./configure
+        --config-cache
+        --with-thread-sanitizer
+        --with-pydebug
+        ${{ fromJSON(inputs.free-threading) && '--disable-gil' || '' }}
     - name: Build CPython
       run: make -j4
     - name: Display build info
@@ -81,6 +81,11 @@ jobs:
       if: always()
       uses: actions/upload-artifact@v4
       with:
-        name: ${{ inputs.tsan_logs_artifact_name }}
+        name: >-
+          tsan-logs-${{
+            fromJSON(inputs.free-threading)
+            && 'free-threading'
+            || 'default'
+          }}
         path: tsan_log.*
         if-no-files-found: ignore

.github/workflows/reusable-windows.yml

@@ -24,7 +24,7 @@ env:
 
 jobs:
   build:
-    name: 'build and test (${{ inputs.arch }})'
+    name: ${{ inputs.arch == 'arm64' && 'build' || 'build and test' }} (${{ inputs.arch }})
     runs-on: ${{ inputs.os }}
     timeout-minutes: 60
     env:

Include/cpython/longintrepr.h

@@ -76,8 +76,8 @@ typedef long stwodigits; /* signed variant of twodigits */
     - 1: Zero
     - 2: Negative
 
-   The third lowest bit of lv_tag is reserved for an immortality flag, but is
-   not currently used.
+   The third lowest bit of lv_tag is
+   set to 1 for the small ints.
 
    In a normalized number, ob_digit[ndigits-1] (the most significant
    digit) is never zero.  Also, in all cases, for all valid i,

Include/cpython/pyatomic.h

@@ -574,15 +574,15 @@ static inline void _Py_atomic_fence_release(void);
 
 #if _Py_USE_GCC_BUILTIN_ATOMICS
 #  define Py_ATOMIC_GCC_H
-#  include "cpython/pyatomic_gcc.h"
+#  include "pyatomic_gcc.h"
 #  undef Py_ATOMIC_GCC_H
 #elif __STDC_VERSION__ >= 201112L && !defined(__STDC_NO_ATOMICS__)
 #  define Py_ATOMIC_STD_H
-#  include "cpython/pyatomic_std.h"
+#  include "pyatomic_std.h"
 #  undef Py_ATOMIC_STD_H
 #elif defined(_MSC_VER)
 #  define Py_ATOMIC_MSC_H
-#  include "cpython/pyatomic_msc.h"
+#  include "pyatomic_msc.h"
 #  undef Py_ATOMIC_MSC_H
 #else
 #  error "no available pyatomic implementation for this platform/compiler"

Include/cpython/pythread.h

@@ -22,7 +22,7 @@ PyAPI_DATA(const long long) PY_TIMEOUT_MAX;
     */
 #   define NATIVE_TSS_KEY_T     unsigned long
 #elif defined(HAVE_PTHREAD_STUBS)
-#   include "cpython/pthread_stubs.h"
+#   include "pthread_stubs.h"
 #   define NATIVE_TSS_KEY_T     pthread_key_t
 #else
 #   error "Require native threads. See https://bugs.python.org/issue31370"

Include/internal/pycore_frame.h

@@ -159,13 +159,6 @@ static inline void _PyFrame_Copy(_PyInterpreterFrame *src, _PyInterpreterFrame *
     // Don't leave a dangling pointer to the old frame when creating generators
     // and coroutines:
     dest->previous = NULL;
-
-#ifdef Py_GIL_DISABLED
-    PyCodeObject *co = _PyFrame_GetCode(dest);
-    for (int i = stacktop; i < co->co_nlocalsplus + co->co_stacksize; i++) {
-        dest->localsplus[i] = PyStackRef_NULL;
-    }
-#endif
 }
 
 #ifdef Py_GIL_DISABLED
@@ -222,16 +215,6 @@ _PyFrame_Initialize(
     for (int i = null_locals_from; i < code->co_nlocalsplus; i++) {
         frame->localsplus[i] = PyStackRef_NULL;
     }
-
-#ifdef Py_GIL_DISABLED
-    // On GIL disabled, we walk the entire stack in GC. Since stacktop
-    // is not always in sync with the real stack pointer, we have
-    // no choice but to traverse the entire stack.
-    // This just makes sure we don't pass the GC invalid stack values.
-    for (int i = code->co_nlocalsplus; i < code->co_nlocalsplus + code->co_stacksize; i++) {
-        frame->localsplus[i] = PyStackRef_NULL;
-    }
-#endif
 }
 
 /* Gets the pointer to the locals array
@@ -405,13 +388,6 @@ _PyFrame_PushTrampolineUnchecked(PyThreadState *tstate, PyCodeObject *code, int
     frame->lltrace = 0;
 #endif
     frame->return_offset = 0;
-
-#ifdef Py_GIL_DISABLED
-    assert(code->co_nlocalsplus == 0);
-    for (int i = 0; i < code->co_stacksize; i++) {
-        frame->localsplus[i] = PyStackRef_NULL;
-    }
-#endif
     return frame;
 }
 

Include/internal/pycore_freelist_state.h

@@ -11,6 +11,8 @@ extern "C" {
 #  define PyTuple_MAXSAVESIZE 20     // Largest tuple to save on freelist
 #  define Py_tuple_MAXFREELIST 2000  // Maximum number of tuples of each size to save
 #  define Py_lists_MAXFREELIST 80
+#  define Py_list_iters_MAXFREELIST 10
+#  define Py_tuple_iters_MAXFREELIST 10
 #  define Py_dicts_MAXFREELIST 80
 #  define Py_dictkeys_MAXFREELIST 80
 #  define Py_floats_MAXFREELIST 100
@@ -40,6 +42,8 @@ struct _Py_freelists {
     struct _Py_freelist ints;
     struct _Py_freelist tuples[PyTuple_MAXSAVESIZE];
     struct _Py_freelist lists;
+    struct _Py_freelist list_iters;
+    struct _Py_freelist tuple_iters;
     struct _Py_freelist dicts;
     struct _Py_freelist dictkeys;
     struct _Py_freelist slices;

Include/internal/pycore_long.h

@@ -159,13 +159,14 @@ PyAPI_FUNC(int) _PyLong_Size_t_Converter(PyObject *, void *);
 
 /* Long value tag bits:
  * 0-1: Sign bits value = (1-sign), ie. negative=2, positive=0, zero=1.
- * 2: Reserved for immortality bit
+ * 2: Set to 1 for the small ints
  * 3+ Unsigned digit count
  */
 #define SIGN_MASK 3
 #define SIGN_ZERO 1
 #define SIGN_NEGATIVE 2
 #define NON_SIZE_BITS 3
+#define IMMORTALITY_BIT_MASK (1 << 2)
 
 /* The functions _PyLong_IsCompact and _PyLong_CompactValue are defined
  * in Include/cpython/longobject.h, since they need to be inline.
@@ -196,7 +197,7 @@ PyAPI_FUNC(int) _PyLong_Size_t_Converter(PyObject *, void *);
 static inline int
 _PyLong_IsNonNegativeCompact(const PyLongObject* op) {
     assert(PyLong_Check(op));
-    return op->long_value.lv_tag <= (1 << NON_SIZE_BITS);
+    return ((op->long_value.lv_tag & ~IMMORTALITY_BIT_MASK) <= (1 << NON_SIZE_BITS));
 }
 
 
@@ -298,7 +299,7 @@ _PyLong_FlipSign(PyLongObject *op) {
         .long_value  = { \
             .lv_tag = TAG_FROM_SIGN_AND_SIZE( \
                 (val) == 0 ? 0 : ((val) < 0 ? -1 : 1), \
-                (val) == 0 ? 0 : 1), \
+                (val) == 0 ? 0 : 1) | IMMORTALITY_BIT_MASK, \
             { ((val) >= 0 ? (val) : -(val)) }, \
         } \
     }