Merge commit 'e933e21e77d4b3c8fe6664b23efe4b0cf151b5da' into pybind11…

…k_merge_sh

.github/workflows/emscripten.yaml

@@ -0,0 +1,30 @@
+name: WASM
+
+on:
+  workflow_dispatch:
+  pull_request:
+    branches:
+    - master
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  build-wasm-emscripten:
+    name: Pyodide wheel
+    runs-on: ubuntu-22.04
+    steps:
+    - uses: actions/checkout@v4
+      with:
+        submodules: true
+        fetch-depth: 0
+
+    - uses: pypa/cibuildwheel@v2.19
+      env:
+        PYODIDE_BUILD_EXPORTS: whole_archive
+        CFLAGS: -fexceptions
+        LDFLAGS: -fexceptions
+      with:
+        package-dir: tests
+        only: cp312-pyodide_wasm32

.pre-commit-config.yaml

@@ -25,22 +25,22 @@ repos:
 
 # Clang format the codebase automatically
 - repo: https://github.com/pre-commit/mirrors-clang-format
-  rev: "v18.1.5"
+  rev: "v18.1.8"
   hooks:
   - id: clang-format
     types_or: [c++, c, cuda]
 
 # Ruff, the Python auto-correcting linter/formatter written in Rust
 - repo: https://github.com/astral-sh/ruff-pre-commit
-  rev: v0.4.7
+  rev: v0.5.0
   hooks:
   - id: ruff
     args: ["--fix", "--show-fixes"]
   - id: ruff-format
 
 # Check static types with mypy
 - repo: https://github.com/pre-commit/mirrors-mypy
-  rev: "v1.10.0"
+  rev: "v1.10.1"
   hooks:
   - id: mypy
     args: []
@@ -80,7 +80,7 @@ repos:
 
 # Also code format the docs
 - repo: https://github.com/adamchainz/blacken-docs
-  rev: "1.16.0"
+  rev: "1.18.0"
   hooks:
   - id: blacken-docs
     additional_dependencies:
@@ -144,14 +144,14 @@ repos:
 
 # PyLint has native support - not always usable, but works for us
 - repo: https://github.com/PyCQA/pylint
-  rev: "v3.2.2"
+  rev: "v3.2.4"
   hooks:
   - id: pylint
     files: ^pybind11
 
 # Check schemas on some of our YAML files
 - repo: https://github.com/python-jsonschema/check-jsonschema
-  rev: 0.28.4
+  rev: 0.28.6
   hooks:
   - id: check-readthedocs
   - id: check-github-workflows

CMakeLists.txt

@@ -164,6 +164,7 @@ set(PYBIND11_HEADERS
     include/pybind11/detail/type_caster_odr_guard.h
     include/pybind11/detail/type_map.h
     include/pybind11/detail/typeid.h
+    include/pybind11/detail/value_and_holder.h
     include/pybind11/attr.h
     include/pybind11/buffer_info.h
     include/pybind11/cast.h

docs/requirements.txt

@@ -16,9 +16,9 @@ breathe==4.35.0 \
     --hash=sha256:5165541c3c67b6c7adde8b3ecfe895c6f7844783c4076b6d8d287e4f33d62386 \
     --hash=sha256:52c581f42ca4310737f9e435e3851c3d1f15446205a85fbc272f1f97ed74f5be
     # via -r requirements.in
-certifi==2024.2.2 \
-    --hash=sha256:0569859f95fc761b18b45ef421b1290a0f65f147e92a1e5eb3e635f9a5e4e66f \
-    --hash=sha256:dc383c07b76109f368f6106eee2b593b04a011ea4d55f652c6ca24a754d1cdd1
+certifi==2024.7.4 \
+    --hash=sha256:5a1e7645bc0ec61a09e26c36f6106dd4cf40c6db3a1fb6352b0244e7fb057c7b \
+    --hash=sha256:c198e21b1289c2ab85ee4e67bb4b4ef3ead0892059901a8d5b622f24a1101e90
     # via requests
 charset-normalizer==3.3.2 \
     --hash=sha256:06435b539f889b1f6f4ac1758871aae42dc3a8c0e24ac9e60c2384973ad73027 \

include/pybind11/cast.h

@@ -952,11 +952,11 @@ struct copyable_holder_caster : public type_caster_base<type> {
         }
     }
 
-    bool load_value(value_and_holder &&v_h) {
+    void load_value(value_and_holder &&v_h) {
         if (v_h.holder_constructed()) {
             value = v_h.value_ptr();
             holder = v_h.template holder<holder_type>();
-            return true;
+            return;
         }
         throw cast_error("Unable to cast from non-held to held instance (T& to Holder<T>) "
 #if !defined(PYBIND11_DETAILED_ERROR_MESSAGES)

include/pybind11/detail/common.h

@@ -462,6 +462,22 @@ PYBIND11_WARNING_POP
                 return "Hello, World!";
             });
         }
+
+    The third macro argument is optional (available since 2.13.0), and can be used to
+    mark the extension module as safe to run without the GIL under a free-threaded CPython
+    interpreter. Passing this argument has no effect on other interpreters.
+
+    .. code-block:: cpp
+
+        PYBIND11_MODULE(example, m, py::mod_gil_not_used()) {
+            m.doc() = "pybind11 example module safe to run without the GIL";
+
+            // Add bindings here
+            m.def("foo", []() {
+                return "Hello, Free-threaded World!";
+            });
+        }
+
 \endrst */
 #define PYBIND11_MODULE(name, variable, ...)                                                      \
     static ::pybind11::module_::module_def PYBIND11_CONCAT(pybind11_module_def_, name)            \

include/pybind11/detail/init.h

@@ -132,7 +132,7 @@ void construct(value_and_holder &v_h, Cpp<Class> *ptr, bool need_alias) {
         // the holder and destruction happens when we leave the C++ scope, and the holder
         // class gets to handle the destruction however it likes.
         v_h.value_ptr() = ptr;
-        v_h.set_instance_registered(true); // SHORTCUT To prevent init_instance from registering it
+        v_h.set_instance_registered(true); // Trick to prevent init_instance from registering it
         // DANGER ZONE BEGIN: exceptions will leave v_h in an invalid state.
         v_h.type->init_instance(v_h.inst, nullptr);                        // Set up the holder
         Holder<Class> temp_holder(std::move(v_h.holder<Holder<Class>>())); // Steal the holder

include/pybind11/detail/internals.h

@@ -92,20 +92,35 @@ struct override_hash {
 
 using instance_map = std::unordered_multimap<const void *, instance *>;
 
+#ifdef Py_GIL_DISABLED
+// Wrapper around PyMutex to provide BasicLockable semantics
+class pymutex {
+    PyMutex mutex;
+
+public:
+    pymutex() : mutex({}) {}
+    void lock() { PyMutex_Lock(&mutex); }
+    void unlock() { PyMutex_Unlock(&mutex); }
+};
+
 // Instance map shards are used to reduce mutex contention in free-threaded Python.
 struct instance_map_shard {
-    std::mutex mutex;
     instance_map registered_instances;
+    pymutex mutex;
     // alignas(64) would be better, but causes compile errors in macOS before 10.14 (see #5200)
-    char padding[64 - (sizeof(std::mutex) + sizeof(instance_map)) % 64];
+    char padding[64 - (sizeof(instance_map) + sizeof(pymutex)) % 64];
 };
 
+static_assert(sizeof(instance_map_shard) % 64 == 0,
+              "instance_map_shard size is not a multiple of 64 bytes");
+#endif
+
 /// Internal data structure used to track registered instances and types.
 /// Whenever binary incompatible changes are made to this structure,
 /// `PYBIND11_INTERNALS_VERSION` must be incremented.
 struct internals {
 #ifdef Py_GIL_DISABLED
-    std::mutex mutex;
+    pymutex mutex;
 #endif
     // std::type_index -> pybind11's type information
     type_map<type_info *> registered_types_cpp;
@@ -467,7 +482,7 @@ inline local_internals &get_local_internals() {
 }
 
 #ifdef Py_GIL_DISABLED
-#    define PYBIND11_LOCK_INTERNALS(internals) std::unique_lock<std::mutex> lock((internals).mutex)
+#    define PYBIND11_LOCK_INTERNALS(internals) std::unique_lock<pymutex> lock((internals).mutex)
 #else
 #    define PYBIND11_LOCK_INTERNALS(internals)
 #endif
@@ -504,7 +519,7 @@ inline auto with_instance_map(const void *ptr,
     auto idx = static_cast<size_t>(hash & internals.instance_shards_mask);
 
     auto &shard = internals.instance_shards[idx];
-    std::unique_lock<std::mutex> lock(shard.mutex);
+    std::unique_lock<pymutex> lock(shard.mutex);
     return cb(shard.registered_instances);
 #else
     (void) ptr;
@@ -520,7 +535,7 @@ inline size_t num_registered_instances() {
     size_t count = 0;
     for (size_t i = 0; i <= internals.instance_shards_mask; ++i) {
         auto &shard = internals.instance_shards[i];
-        std::unique_lock<std::mutex> lock(shard.mutex);
+        std::unique_lock<pymutex> lock(shard.mutex);
         count += shard.registered_instances.size();
     }
     return count;

include/pybind11/detail/type_caster_base.h

@@ -14,6 +14,7 @@
 #include "descr.h"
 #include "internals.h"
 #include "typeid.h"
+#include "value_and_holder.h"
 
 #include <cstdint>
 #include <iterator>
@@ -259,67 +260,6 @@ PYBIND11_NOINLINE handle find_registered_python_instance(void *src,
     });
 }
 
-struct value_and_holder {
-    instance *inst = nullptr;
-    size_t index = 0u;
-    const detail::type_info *type = nullptr;
-    void **vh = nullptr;
-
-    // Main constructor for a found value/holder:
-    value_and_holder(instance *i, const detail::type_info *type, size_t vpos, size_t index)
-        : inst{i}, index{index}, type{type},
-          vh{inst->simple_layout ? inst->simple_value_holder
-                                 : &inst->nonsimple.values_and_holders[vpos]} {}
-
-    // Default constructor (used to signal a value-and-holder not found by get_value_and_holder())
-    value_and_holder() = default;
-
-    // Used for past-the-end iterator
-    explicit value_and_holder(size_t index) : index{index} {}
-
-    template <typename V = void>
-    V *&value_ptr() const {
-        return reinterpret_cast<V *&>(vh[0]);
-    }
-    // True if this `value_and_holder` has a non-null value pointer
-    explicit operator bool() const { return value_ptr() != nullptr; }
-
-    template <typename H>
-    H &holder() const {
-        return reinterpret_cast<H &>(vh[1]);
-    }
-    bool holder_constructed() const {
-        return inst->simple_layout
-                   ? inst->simple_holder_constructed
-                   : (inst->nonsimple.status[index] & instance::status_holder_constructed) != 0u;
-    }
-    // NOLINTNEXTLINE(readability-make-member-function-const)
-    void set_holder_constructed(bool v = true) {
-        if (inst->simple_layout) {
-            inst->simple_holder_constructed = v;
-        } else if (v) {
-            inst->nonsimple.status[index] |= instance::status_holder_constructed;
-        } else {
-            inst->nonsimple.status[index] &= (std::uint8_t) ~instance::status_holder_constructed;
-        }
-    }
-    bool instance_registered() const {
-        return inst->simple_layout
-                   ? inst->simple_instance_registered
-                   : ((inst->nonsimple.status[index] & instance::status_instance_registered) != 0);
-    }
-    // NOLINTNEXTLINE(readability-make-member-function-const)
-    void set_instance_registered(bool v = true) {
-        if (inst->simple_layout) {
-            inst->simple_instance_registered = v;
-        } else if (v) {
-            inst->nonsimple.status[index] |= instance::status_instance_registered;
-        } else {
-            inst->nonsimple.status[index] &= (std::uint8_t) ~instance::status_instance_registered;
-        }
-    }
-};
-
 // Container for accessing and iterating over an instance's values/holders
 struct values_and_holders {
 private:

include/pybind11/detail/value_and_holder.h

@@ -0,0 +1,77 @@
+// Copyright (c) 2016-2024 The Pybind Development Team.
+// All rights reserved. Use of this source code is governed by a
+// BSD-style license that can be found in the LICENSE file.
+
+#pragma once
+
+#include "common.h"
+
+#include <cstddef>
+#include <typeinfo>
+
+PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)
+PYBIND11_NAMESPACE_BEGIN(detail)
+
+struct value_and_holder {
+    instance *inst = nullptr;
+    size_t index = 0u;
+    const detail::type_info *type = nullptr;
+    void **vh = nullptr;
+
+    // Main constructor for a found value/holder:
+    value_and_holder(instance *i, const detail::type_info *type, size_t vpos, size_t index)
+        : inst{i}, index{index}, type{type},
+          vh{inst->simple_layout ? inst->simple_value_holder
+                                 : &inst->nonsimple.values_and_holders[vpos]} {}
+
+    // Default constructor (used to signal a value-and-holder not found by get_value_and_holder())
+    value_and_holder() = default;
+
+    // Used for past-the-end iterator
+    explicit value_and_holder(size_t index) : index{index} {}
+
+    template <typename V = void>
+    V *&value_ptr() const {
+        return reinterpret_cast<V *&>(vh[0]);
+    }
+    // True if this `value_and_holder` has a non-null value pointer
+    explicit operator bool() const { return value_ptr() != nullptr; }
+
+    template <typename H>
+    H &holder() const {
+        return reinterpret_cast<H &>(vh[1]);
+    }
+    bool holder_constructed() const {
+        return inst->simple_layout
+                   ? inst->simple_holder_constructed
+                   : (inst->nonsimple.status[index] & instance::status_holder_constructed) != 0u;
+    }
+    // NOLINTNEXTLINE(readability-make-member-function-const)
+    void set_holder_constructed(bool v = true) {
+        if (inst->simple_layout) {
+            inst->simple_holder_constructed = v;
+        } else if (v) {
+            inst->nonsimple.status[index] |= instance::status_holder_constructed;
+        } else {
+            inst->nonsimple.status[index] &= (std::uint8_t) ~instance::status_holder_constructed;
+        }
+    }
+    bool instance_registered() const {
+        return inst->simple_layout
+                   ? inst->simple_instance_registered
+                   : ((inst->nonsimple.status[index] & instance::status_instance_registered) != 0);
+    }
+    // NOLINTNEXTLINE(readability-make-member-function-const)
+    void set_instance_registered(bool v = true) {
+        if (inst->simple_layout) {
+            inst->simple_instance_registered = v;
+        } else if (v) {
+            inst->nonsimple.status[index] |= instance::status_instance_registered;
+        } else {
+            inst->nonsimple.status[index] &= (std::uint8_t) ~instance::status_instance_registered;
+        }
+    }
+};
+
+PYBIND11_NAMESPACE_END(detail)
+PYBIND11_NAMESPACE_END(PYBIND11_NAMESPACE)

include/pybind11/gil_safe_call_once.h

@@ -8,6 +8,10 @@
 #include <cassert>
 #include <mutex>
 
+#ifdef Py_GIL_DISABLED
+#    include <atomic>
+#endif
+
 PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)
 
 // Use the `gil_safe_call_once_and_store` class below instead of the naive
@@ -82,7 +86,12 @@ class gil_safe_call_once_and_store {
 private:
     alignas(T) char storage_[sizeof(T)] = {};
     std::once_flag once_flag_ = {};
-    bool is_initialized_ = false;
+#ifdef Py_GIL_DISABLED
+    std::atomic_bool
+#else
+    bool
+#endif
+        is_initialized_{false};
     // The `is_initialized_`-`storage_` pair is very similar to `std::optional`,
     // but the latter does not have the triviality properties of former,
     // therefore `std::optional` is not a viable alternative here.