Merge branch 'master' into sh_merge_master

.pre-commit-config.yaml

@@ -25,14 +25,14 @@ repos:
 
 # Clang format the codebase automatically
 - repo: https://github.com/pre-commit/mirrors-clang-format
-  rev: "v18.1.8"
+  rev: "v19.1.3"
   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: v0.6.3
+  rev: v0.7.2
   hooks:
   - id: ruff
     args: ["--fix", "--show-fixes"]
@@ -40,7 +40,7 @@ repos:
 
 # Check static types with mypy
 - repo: https://github.com/pre-commit/mirrors-mypy
-  rev: "v1.11.2"
+  rev: "v1.13.0"
   hooks:
   - id: mypy
     args: []
@@ -62,7 +62,7 @@ repos:
 
 # Standard hooks
 - repo: https://github.com/pre-commit/pre-commit-hooks
-  rev: "v4.6.0"
+  rev: "v5.0.0"
   hooks:
   - id: check-added-large-files
   - id: check-case-conflict
@@ -80,7 +80,7 @@ repos:
 
 # Also code format the docs
 - repo: https://github.com/adamchainz/blacken-docs
-  rev: "1.18.0"
+  rev: "1.19.1"
   hooks:
   - id: blacken-docs
     additional_dependencies:
@@ -110,7 +110,7 @@ repos:
 
 # Checks the manifest for missing files (native support)
 - repo: https://github.com/mgedmin/check-manifest
-  rev: "0.49"
+  rev: "0.50"
   hooks:
   - id: check-manifest
     # This is a slow hook, so only run this if --hook-stage manual is passed
@@ -144,14 +144,14 @@ repos:
 
 # PyLint has native support - not always usable, but works for us
 - repo: https://github.com/PyCQA/pylint
-  rev: "v3.2.7"
+  rev: "v3.3.1"
   hooks:
   - id: pylint
     files: ^pybind11
 
 # Check schemas on some of our YAML files
 - repo: https://github.com/python-jsonschema/check-jsonschema
-  rev: 0.29.2
+  rev: 0.29.4
   hooks:
   - id: check-readthedocs
   - id: check-github-workflows

docs/upgrade.rst

@@ -24,7 +24,8 @@ changes are that:
   function is not available anymore.
 
 Due to NumPy changes, you may experience difficulties updating to NumPy 2.
-Please see the [NumPy 2 migration guide](https://numpy.org/devdocs/numpy_2_0_migration_guide.html) for details.
+Please see the `NumPy 2 migration guide <https://numpy.org/devdocs/numpy_2_0_migration_guide.html>`_
+for details.
 For example, a more direct change could be that the default integer ``"int_"``
 (and ``"uint"``) is now ``ssize_t`` and not ``long`` (affects 64bit windows).
 

include/pybind11/detail/common.h

@@ -1150,14 +1150,14 @@ struct overload_cast_impl {
     }
 
     template <typename Return, typename Class>
-    constexpr auto operator()(Return (Class::*pmf)(Args...),
+    constexpr auto operator()(Return (Class::*pmf)(Args...), std::false_type = {}) const noexcept
-                              std::false_type = {}) const noexcept -> decltype(pmf) {
+        -> decltype(pmf) {
         return pmf;
     }
 
     template <typename Return, typename Class>
-    constexpr auto operator()(Return (Class::*pmf)(Args...) const,
+    constexpr auto operator()(Return (Class::*pmf)(Args...) const, std::true_type) const noexcept
-                              std::true_type) const noexcept -> decltype(pmf) {
+        -> decltype(pmf) {
         return pmf;
     }
 };

include/pybind11/detail/descr.h

@@ -156,9 +156,8 @@ constexpr auto concat(const descr<N, Ts...> &d, const Args &...args) {
 }
 #else
 template <size_t N, typename... Ts, typename... Args>
-constexpr auto concat(const descr<N, Ts...> &d,
+constexpr auto concat(const descr<N, Ts...> &d, const Args &...args)
-                      const Args &...args) -> decltype(std::declval<descr<N + 2, Ts...>>()
+    -> decltype(std::declval<descr<N + 2, Ts...>>() + concat(args...)) {
-                                                       + concat(args...)) {
     return d + const_name(", ") + concat(args...);
 }
 #endif

include/pybind11/detail/internals.h

@@ -697,8 +697,8 @@ inline std::uint64_t mix64(std::uint64_t z) {
 }
 
 template <typename F>
-inline auto with_instance_map(const void *ptr,
+inline auto with_instance_map(const void *ptr, const F &cb)
-                              const F &cb) -> decltype(cb(std::declval<instance_map &>())) {
+    -> decltype(cb(std::declval<instance_map &>())) {
     auto &internals = get_internals();
 
 #ifdef Py_GIL_DISABLED

include/pybind11/detail/type_caster_base.h

@@ -121,7 +121,6 @@ PYBIND11_NOINLINE void all_type_info_populate(PyTypeObject *t, std::vector<type_
     for (handle parent : reinterpret_borrow<tuple>(t->tp_bases)) {
         check.push_back((PyTypeObject *) parent.ptr());
     }
-
     auto const &type_dict = get_internals().registered_types_py;
     for (size_t i = 0; i < check.size(); i++) {
         auto *type = check[i];
@@ -180,13 +179,7 @@ PYBIND11_NOINLINE void all_type_info_populate(PyTypeObject *t, std::vector<type_
  * The value is cached for the lifetime of the Python type.
  */
 inline const std::vector<detail::type_info *> &all_type_info(PyTypeObject *type) {
-    auto ins = all_type_info_get_cache(type);
+    return all_type_info_get_cache(type).first->second;
-    if (ins.second) {
-        // New cache entry: populate it
-        all_type_info_populate(type, ins.first->second);
-    }
-
-    return ins.first->second;
 }
 
 /**
@@ -1535,14 +1528,14 @@ protected:
        does not have a private operator new implementation. A comma operator is used in the
        decltype argument to apply SFINAE to the public copy/move constructors.*/
     template <typename T, typename = enable_if_t<is_copy_constructible<T>::value>>
-    static auto make_copy_constructor(const T *) -> decltype(new T(std::declval<const T>()),
+    static auto make_copy_constructor(const T *)
-                                                             Constructor{}) {
+        -> decltype(new T(std::declval<const T>()), Constructor{}) {
         return [](const void *arg) -> void * { return new T(*reinterpret_cast<const T *>(arg)); };
     }
 
     template <typename T, typename = enable_if_t<is_move_constructible<T>::value>>
-    static auto make_move_constructor(const T *) -> decltype(new T(std::declval<T &&>()),
+    static auto make_move_constructor(const T *)
-                                                             Constructor{}) {
+        -> decltype(new T(std::declval<T &&>()), Constructor{}) {
         return [](const void *arg) -> void * {
             return new T(std::move(*const_cast<T *>(reinterpret_cast<const T *>(arg))));
         };

include/pybind11/eigen/tensor.h

@@ -124,9 +124,9 @@ struct eigen_tensor_helper<
 template <typename Type, bool ShowDetails, bool NeedsWriteable = false>
 struct get_tensor_descriptor {
     static constexpr auto details
-        = const_name<NeedsWriteable>(", flags.writeable", "")
+        = const_name<NeedsWriteable>(", flags.writeable", "") + const_name
-          + const_name<static_cast<int>(Type::Layout) == static_cast<int>(Eigen::RowMajor)>(
+              < static_cast<int>(Type::Layout)
-              ", flags.c_contiguous", ", flags.f_contiguous");
+          == static_cast<int>(Eigen::RowMajor) > (", flags.c_contiguous", ", flags.f_contiguous");
     static constexpr auto value
         = const_name("numpy.ndarray[") + npy_format_descriptor<typename Type::Scalar>::name
           + const_name("[") + eigen_tensor_helper<remove_cv_t<Type>>::dimensions_descriptor

include/pybind11/pybind11.h

@@ -2665,13 +2665,20 @@ keep_alive_impl(size_t Nurse, size_t Patient, function_call &call, handle ret) {
 inline std::pair<decltype(internals::registered_types_py)::iterator, bool>
 all_type_info_get_cache(PyTypeObject *type) {
     auto res = with_internals([type](internals &internals) {
-        return internals
+        auto ins = internals
                        .registered_types_py
 #ifdef __cpp_lib_unordered_map_try_emplace
                        .try_emplace(type);
 #else
                        .emplace(type, std::vector<detail::type_info *>());
 #endif
+        if (ins.second) {
+            // For free-threading mode, this call must be under
+            // the with_internals() mutex lock, to avoid that other threads
+            // continue running with the empty ins.first->second.
+            all_type_info_populate(type, ins.first->second);
+        }
+        return ins;
     });
     if (res.second) {
         // New cache entry created; set up a weak reference to automatically remove it if the type

tests/pybind11_tests.cpp

@@ -128,4 +128,9 @@ PYBIND11_MODULE(pybind11_tests, m, py::mod_gil_not_used()) {
     for (const auto &initializer : initializers()) {
         initializer(m);
     }
+
+    py::class_<TestContext>(m, "TestContext")
+        .def(py::init<>(&TestContext::createNewContextForInit))
+        .def("__enter__", &TestContext::contextEnter)
+        .def("__exit__", &TestContext::contextExit);
 }

tests/pybind11_tests.h

@@ -96,3 +96,24 @@ void ignoreOldStyleInitWarnings(F &&body) {
     )",
              py::dict(py::arg("body") = py::cpp_function(body)));
 }
+
+// See PR #5419 for background.
+class TestContext {
+public:
+    TestContext() = delete;
+    TestContext(const TestContext &) = delete;
+    TestContext(TestContext &&) = delete;
+    static TestContext *createNewContextForInit() { return new TestContext("new-context"); }
+
+    pybind11::object contextEnter() {
+        py::object contextObj = py::cast(*this);
+        return contextObj;
+    }
+    void contextExit(const pybind11::object & /*excType*/,
+                     const pybind11::object & /*excVal*/,
+                     const pybind11::object & /*excTb*/) {}
+
+private:
+    explicit TestContext(const std::string &context) : context(context) {}
+    std::string context;
+};

tests/test_class.py

@@ -1,5 +1,6 @@
 from __future__ import annotations
 
+import sys
 from unittest import mock
 
 import pytest
@@ -520,3 +521,31 @@ def test_pr4220_tripped_over_this():
         m.Empty0().get_msg()
         == "This is really only meant to exercise successful compilation."
     )
+
+
+@pytest.mark.skipif(sys.platform.startswith("emscripten"), reason="Requires threads")
+def test_all_type_info_multithreaded():
+    # See PR #5419 for background.
+    import threading
+
+    from pybind11_tests import TestContext
+
+    class Context(TestContext):
+        pass
+
+    num_runs = 10
+    num_threads = 4
+    barrier = threading.Barrier(num_threads)
+
+    def func():
+        barrier.wait()
+        with Context():
+            pass
+
+    for _ in range(num_runs):
+        threads = [threading.Thread(target=func) for _ in range(num_threads)]
+        for thread in threads:
+            thread.start()
+
+        for thread in threads:
+            thread.join()