diff --git a/include/pybind11/attr.h b/include/pybind11/attr.h
index 38139afa32..1044db94d9 100644
--- a/include/pybind11/attr.h
+++ b/include/pybind11/attr.h
@@ -10,6 +10,7 @@
 
 #pragma once
 
+#include "detail/common.h"
 #include "cast.h"
 
 #include <functional>
@@ -25,6 +26,9 @@ struct is_method {
     explicit is_method(const handle &c) : class_(c) {}
 };
 
+/// Annotation for setters
+struct is_setter {};
+
 /// Annotation for operators
 struct is_operator {};
 
@@ -187,8 +191,8 @@ struct argument_record {
 struct function_record {
     function_record()
         : is_constructor(false), is_new_style_constructor(false), is_stateless(false),
-          is_operator(false), is_method(false), has_args(false), has_kwargs(false),
-          prepend(false) {}
+          is_operator(false), is_method(false), is_setter(false), has_args(false),
+          has_kwargs(false), prepend(false) {}
 
     /// Function name
     char *name = nullptr; /* why no C++ strings? They generate heavier code.. */
@@ -229,6 +233,9 @@ struct function_record {
     /// True if this is a method
     bool is_method : 1;
 
+    /// True if this is a setter
+    bool is_setter : 1;
+
     /// True if the function has a '*args' argument
     bool has_args : 1;
 
@@ -344,9 +351,11 @@ struct type_record {
 
         bases.append((PyObject *) base_info->type);
 
-        if (base_info->type->tp_dictoffset != 0) {
-            dynamic_attr = true;
-        }
+#if PY_VERSION_HEX < 0x030B0000
+        dynamic_attr |= base_info->type->tp_dictoffset != 0;
+#else
+        dynamic_attr |= (base_info->type->tp_flags & Py_TPFLAGS_MANAGED_DICT) != 0;
+#endif
 
         if (caster) {
             base_info->implicit_casts.emplace_back(type, caster);
@@ -396,7 +405,7 @@ struct process_attribute<doc> : process_attribute_default<doc> {
 template <>
 struct process_attribute<const char *> : process_attribute_default<const char *> {
     static void init(const char *d, function_record *r) { r->doc = const_cast<char *>(d); }
-    static void init(const char *d, type_record *r) { r->doc = const_cast<char *>(d); }
+    static void init(const char *d, type_record *r) { r->doc = d; }
 };
 template <>
 struct process_attribute<char *> : process_attribute<const char *> {};
@@ -423,6 +432,12 @@ struct process_attribute<is_method> : process_attribute_default<is_method> {
     }
 };
 
+/// Process an attribute which indicates that this function is a setter
+template <>
+struct process_attribute<is_setter> : process_attribute_default<is_setter> {
+    static void init(const is_setter &, function_record *r) { r->is_setter = true; }
+};
+
 /// Process an attribute which indicates the parent scope of a method
 template <>
 struct process_attribute<scope> : process_attribute_default<scope> {
@@ -477,7 +492,7 @@ struct process_attribute<arg_v> : process_attribute_default<arg_v> {
         }
 
         if (!a.value) {
-#if !defined(NDEBUG)
+#if defined(PYBIND11_DETAILED_ERROR_MESSAGES)
             std::string descr("'");
             if (a.name) {
                 descr += std::string(a.name) + ": ";
@@ -498,7 +513,8 @@ struct process_attribute<arg_v> : process_attribute_default<arg_v> {
 #else
             pybind11_fail("arg(): could not convert default argument "
                           "into a Python object (type not registered yet?). "
-                          "Compile in debug mode for more information.");
+                          "#define PYBIND11_DETAILED_ERROR_MESSAGES or compile in debug mode for "
+                          "more information.");
 #endif
         }
         r->args.emplace_back(a.name, a.descr, a.value.inc_ref(), !a.flag_noconvert, a.flag_none);
diff --git a/include/pybind11/buffer_info.h b/include/pybind11/buffer_info.h
index 06120d5563..75aec0ba30 100644
--- a/include/pybind11/buffer_info.h
+++ b/include/pybind11/buffer_info.h
@@ -37,6 +37,9 @@ inline std::vector<ssize_t> f_strides(const std::vector<ssize_t> &shape, ssize_t
     return strides;
 }
 
+template <typename T, typename SFINAE = void>
+struct compare_buffer_info;
+
 PYBIND11_NAMESPACE_END(detail)
 
 /// Information record describing a Python buffer object
@@ -99,22 +102,22 @@ struct buffer_info {
     template <typename T>
     buffer_info(const T *ptr, ssize_t size, bool readonly = true)
         : buffer_info(
-            const_cast<T *>(ptr), sizeof(T), format_descriptor<T>::format(), size, readonly) {}
+              const_cast<T *>(ptr), sizeof(T), format_descriptor<T>::format(), size, readonly) {}
 
     explicit buffer_info(Py_buffer *view, bool ownview = true)
         : buffer_info(
-            view->buf,
-            view->itemsize,
-            view->format,
-            view->ndim,
-            {view->shape, view->shape + view->ndim},
-            /* Though buffer::request() requests PyBUF_STRIDES, ctypes objects
-             * ignore this flag and return a view with NULL strides.
-             * When strides are NULL, build them manually.  */
-            view->strides
-                ? std::vector<ssize_t>(view->strides, view->strides + view->ndim)
-                : detail::c_strides({view->shape, view->shape + view->ndim}, view->itemsize),
-            (view->readonly != 0)) {
+              view->buf,
+              view->itemsize,
+              view->format,
+              view->ndim,
+              {view->shape, view->shape + view->ndim},
+              /* Though buffer::request() requests PyBUF_STRIDES, ctypes objects
+               * ignore this flag and return a view with NULL strides.
+               * When strides are NULL, build them manually.  */
+              view->strides
+                  ? std::vector<ssize_t>(view->strides, view->strides + view->ndim)
+                  : detail::c_strides({view->shape, view->shape + view->ndim}, view->itemsize),
+              (view->readonly != 0)) {
         // NOLINTNEXTLINE(cppcoreguidelines-prefer-member-initializer)
         this->m_view = view;
         // NOLINTNEXTLINE(cppcoreguidelines-prefer-member-initializer)
@@ -150,6 +153,17 @@ struct buffer_info {
     Py_buffer *view() const { return m_view; }
     Py_buffer *&view() { return m_view; }
 
+    /* True if the buffer item type is equivalent to `T`. */
+    // To define "equivalent" by example:
+    // `buffer_info::item_type_is_equivalent_to<int>(b)` and
+    // `buffer_info::item_type_is_equivalent_to<long>(b)` may both be true
+    // on some platforms, but `int` and `unsigned` will never be equivalent.
+    // For the ground truth, please inspect `detail::compare_buffer_info<>`.
+    template <typename T>
+    bool item_type_is_equivalent_to() const {
+        return detail::compare_buffer_info<T>::compare(*this);
+    }
+
 private:
     struct private_ctr_tag {};
 
@@ -162,7 +176,7 @@ struct buffer_info {
                 detail::any_container<ssize_t> &&strides_in,
                 bool readonly)
         : buffer_info(
-            ptr, itemsize, format, ndim, std::move(shape_in), std::move(strides_in), readonly) {}
+              ptr, itemsize, format, ndim, std::move(shape_in), std::move(strides_in), readonly) {}
 
     Py_buffer *m_view = nullptr;
     bool ownview = false;
@@ -170,9 +184,10 @@ struct buffer_info {
 
 PYBIND11_NAMESPACE_BEGIN(detail)
 
-template <typename T, typename SFINAE = void>
+template <typename T, typename SFINAE>
 struct compare_buffer_info {
     static bool compare(const buffer_info &b) {
+        // NOLINTNEXTLINE(bugprone-sizeof-expression) Needed for `PyObject *`
         return b.format == format_descriptor<T>::format() && b.itemsize == (ssize_t) sizeof(T);
     }
 };
diff --git a/include/pybind11/cast.h b/include/pybind11/cast.h
index d45b49c52e..e41ad2abfa 100644
--- a/include/pybind11/cast.h
+++ b/include/pybind11/cast.h
@@ -29,6 +29,9 @@
 #include <vector>
 
 PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)
+
+PYBIND11_WARNING_DISABLE_MSVC(4127)
+
 PYBIND11_NAMESPACE_BEGIN(detail)
 
 template <typename type, typename SFINAE = void>
@@ -39,13 +42,15 @@ using make_caster = type_caster<intrinsic_t<type>>;
 // Shortcut for calling a caster's `cast_op_type` cast operator for casting a type_caster to a T
 template <typename T>
 typename make_caster<T>::template cast_op_type<T> cast_op(make_caster<T> &caster) {
-    return caster.operator typename make_caster<T>::template cast_op_type<T>();
+    using result_t = typename make_caster<T>::template cast_op_type<T>; // See PR #4893
+    return caster.operator result_t();
 }
 template <typename T>
 typename make_caster<T>::template cast_op_type<typename std::add_rvalue_reference<T>::type>
 cast_op(make_caster<T> &&caster) {
-    return std::move(caster).operator typename make_caster<T>::
-        template cast_op_type<typename std::add_rvalue_reference<T>::type>();
+    using result_t = typename make_caster<T>::template cast_op_type<
+        typename std::add_rvalue_reference<T>::type>; // See PR #4893
+    return std::move(caster).operator result_t();
 }
 
 template <typename type>
@@ -88,7 +93,8 @@ public:
     template <typename T_,                                                                        \
               ::pybind11::detail::enable_if_t<                                                    \
                   std::is_same<type, ::pybind11::detail::remove_cv_t<T_>>::value,                 \
-                  int> = 0>                                                                       \
+                  int>                                                                            \
+              = 0>                                                                                \
     static ::pybind11::handle cast(                                                               \
         T_ *src, ::pybind11::return_value_policy policy, ::pybind11::handle parent) {             \
         if (!src)                                                                                 \
@@ -248,7 +254,7 @@ struct void_caster {
         return false;
     }
     static handle cast(T, return_value_policy /* policy */, handle /* parent */) {
-        return none().inc_ref();
+        return none().release();
     }
     PYBIND11_TYPE_CASTER(T, const_name("None"));
 };
@@ -291,7 +297,7 @@ class type_caster<void> : public type_caster<void_type> {
         if (ptr) {
             return capsule(ptr).release();
         }
-        return none().inc_ref();
+        return none().release();
     }
 
     template <typename T>
@@ -321,8 +327,9 @@ class type_caster<bool> {
             value = false;
             return true;
         }
-        if (convert || (std::strcmp("numpy.bool_", Py_TYPE(src.ptr())->tp_name) == 0)) {
-            // (allow non-implicit conversion for numpy booleans)
+        if (convert || is_numpy_bool(src)) {
+            // (allow non-implicit conversion for numpy booleans), use strncmp
+            // since NumPy 1.x had an additional trailing underscore.
 
             Py_ssize_t res = -1;
             if (src.is_none()) {
@@ -354,6 +361,15 @@ class type_caster<bool> {
         return handle(src ? Py_True : Py_False).inc_ref();
     }
     PYBIND11_TYPE_CASTER(bool, const_name("bool"));
+
+private:
+    // Test if an object is a NumPy boolean (without fetching the type).
+    static inline bool is_numpy_bool(handle object) {
+        const char *type_name = Py_TYPE(object.ptr())->tp_name;
+        // Name changed to `numpy.bool` in NumPy 2, `numpy.bool_` is needed for 1.x support
+        return std::strcmp("numpy.bool", type_name) == 0
+               || std::strcmp("numpy.bool_", type_name) == 0;
+    }
 };
 
 // Helper class for UTF-{8,16,32} C++ stl strings:
@@ -389,7 +405,7 @@ struct string_caster {
 
         // For UTF-8 we avoid the need for a temporary `bytes` object by using
         // `PyUnicode_AsUTF8AndSize`.
-        if (PYBIND11_SILENCE_MSVC_C4127(UTF_N == 8)) {
+        if (UTF_N == 8) {
             Py_ssize_t size = -1;
             const auto *buffer
                 = reinterpret_cast<const CharT *>(PyUnicode_AsUTF8AndSize(load_src.ptr(), &size));
@@ -416,7 +432,7 @@ struct string_caster {
             = reinterpret_cast<const CharT *>(PYBIND11_BYTES_AS_STRING(utfNbytes.ptr()));
         size_t length = (size_t) PYBIND11_BYTES_SIZE(utfNbytes.ptr()) / sizeof(CharT);
         // Skip BOM for UTF-16/32
-        if (PYBIND11_SILENCE_MSVC_C4127(UTF_N > 8)) {
+        if (UTF_N > 8) {
             buffer++;
             length--;
         }
@@ -514,7 +530,7 @@ struct type_caster<std::basic_string_view<CharT, Traits>,
 template <typename CharT>
 struct type_caster<CharT, enable_if_t<is_std_char_type<CharT>::value>> {
     using StringType = std::basic_string<CharT>;
-    using StringCaster = type_caster<StringType>;
+    using StringCaster = make_caster<StringType>;
     StringCaster str_caster;
     bool none = false;
     CharT one_char = 0;
@@ -537,7 +553,7 @@ struct type_caster<CharT, enable_if_t<is_std_char_type<CharT>::value>> {
 
     static handle cast(const CharT *src, return_value_policy policy, handle parent) {
         if (src == nullptr) {
-            return pybind11::none().inc_ref();
+            return pybind11::none().release();
         }
         return StringCaster::cast(StringType(src), policy, parent);
     }
@@ -572,7 +588,7 @@ struct type_caster<CharT, enable_if_t<is_std_char_type<CharT>::value>> {
         // figure out how long the first encoded character is in bytes to distinguish between these
         // two errors.  We also allow want to allow unicode characters U+0080 through U+00FF, as
         // those can fit into a single char value.
-        if (PYBIND11_SILENCE_MSVC_C4127(StringCaster::UTF_N == 8) && str_len > 1 && str_len <= 4) {
+        if (StringCaster::UTF_N == 8 && str_len > 1 && str_len <= 4) {
             auto v0 = static_cast<unsigned char>(value[0]);
             // low bits only: 0-127
             // 0b110xxxxx - start of 2-byte sequence
@@ -598,7 +614,7 @@ struct type_caster<CharT, enable_if_t<is_std_char_type<CharT>::value>> {
         // UTF-16 is much easier: we can only have a surrogate pair for values above U+FFFF, thus a
         // surrogate pair with total length 2 instantly indicates a range error (but not a "your
         // string was too long" error).
-        else if (PYBIND11_SILENCE_MSVC_C4127(StringCaster::UTF_N == 16) && str_len == 2) {
+        else if (StringCaster::UTF_N == 16 && str_len == 2) {
             one_char = static_cast<CharT>(value[0]);
             if (one_char >= 0xD800 && one_char < 0xE000) {
                 throw value_error("Character code point not in range(0x10000)");
@@ -656,8 +672,9 @@ class tuple_caster {
         return cast(*src, policy, parent);
     }
 
-    static constexpr auto name
-        = const_name("Tuple[") + concat(make_caster<Ts>::name...) + const_name("]");
+    static constexpr auto name = const_name("tuple[")
+                                 + ::pybind11::detail::concat(make_caster<Ts>::name...)
+                                 + const_name("]");
 
     template <typename T>
     using cast_op_type = type;
@@ -723,6 +740,13 @@ class type_caster<std::pair<T1, T2>> : public tuple_caster<std::pair, T1, T2> {}
 template <typename... Ts>
 class type_caster<std::tuple<Ts...>> : public tuple_caster<std::tuple, Ts...> {};
 
+template <>
+class type_caster<std::tuple<>> : public tuple_caster<std::tuple> {
+public:
+    // PEP 484 specifies this syntax for an empty tuple
+    static constexpr auto name = const_name("tuple[()]");
+};
+
 /// Helper class which abstracts away certain actions. Users can provide specializations for
 /// custom holders, but it's only necessary if the type has a non-standard interface.
 template <typename T>
@@ -777,8 +801,9 @@ struct copyable_holder_caster : public type_caster_base<type> {
             return true;
         }
         throw cast_error("Unable to cast from non-held to held instance (T& to Holder<T>) "
-#if defined(NDEBUG)
-                         "(compile in debug mode for type information)");
+#if !defined(PYBIND11_DETAILED_ERROR_MESSAGES)
+                         "(#define PYBIND11_DETAILED_ERROR_MESSAGES or compile in debug mode for "
+                         "type information)");
 #else
                          "of type '"
                          + type_id<holder_type>() + "''");
@@ -845,7 +870,7 @@ struct always_construct_holder {
 
 /// Create a specialization for custom holder types (silently ignores std::shared_ptr)
 #define PYBIND11_DECLARE_HOLDER_TYPE(type, holder_type, ...)                                      \
-    namespace pybind11 {                                                                          \
+    PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)                                                  \
     namespace detail {                                                                            \
     template <typename type>                                                                      \
     struct always_construct_holder<holder_type> : always_construct_holder<void, ##__VA_ARGS__> {  \
@@ -854,7 +879,7 @@ struct always_construct_holder {
     class type_caster<holder_type, enable_if_t<!is_shared_ptr<holder_type>::value>>               \
         : public type_caster_holder<type, holder_type> {};                                        \
     }                                                                                             \
-    }
+    PYBIND11_NAMESPACE_END(PYBIND11_NAMESPACE)
 
 // PYBIND11_DECLARE_HOLDER_TYPE holder types:
 template <typename base, typename holder>
@@ -864,10 +889,53 @@ struct is_holder_type
 template <typename base, typename deleter>
 struct is_holder_type<base, std::unique_ptr<base, deleter>> : std::true_type {};
 
+#ifdef PYBIND11_DISABLE_HANDLE_TYPE_NAME_DEFAULT_IMPLEMENTATION // See PR #4888
+
+// This leads to compilation errors if a specialization is missing.
+template <typename T>
+struct handle_type_name;
+
+#else
+
 template <typename T>
 struct handle_type_name {
     static constexpr auto name = const_name<T>();
 };
+
+#endif
+
+template <>
+struct handle_type_name<object> {
+    static constexpr auto name = const_name("object");
+};
+template <>
+struct handle_type_name<list> {
+    static constexpr auto name = const_name("list");
+};
+template <>
+struct handle_type_name<dict> {
+    static constexpr auto name = const_name("dict");
+};
+template <>
+struct handle_type_name<anyset> {
+    static constexpr auto name = const_name("Union[set, frozenset]");
+};
+template <>
+struct handle_type_name<set> {
+    static constexpr auto name = const_name("set");
+};
+template <>
+struct handle_type_name<frozenset> {
+    static constexpr auto name = const_name("frozenset");
+};
+template <>
+struct handle_type_name<str> {
+    static constexpr auto name = const_name("str");
+};
+template <>
+struct handle_type_name<tuple> {
+    static constexpr auto name = const_name("tuple");
+};
 template <>
 struct handle_type_name<bool_> {
     static constexpr auto name = const_name("bool");
@@ -877,6 +945,10 @@ struct handle_type_name<bytes> {
     static constexpr auto name = const_name(PYBIND11_BYTES_NAME);
 };
 template <>
+struct handle_type_name<buffer> {
+    static constexpr auto name = const_name("Buffer");
+};
+template <>
 struct handle_type_name<int_> {
     static constexpr auto name = const_name("int");
 };
@@ -893,10 +965,50 @@ struct handle_type_name<float_> {
     static constexpr auto name = const_name("float");
 };
 template <>
+struct handle_type_name<function> {
+    static constexpr auto name = const_name("Callable");
+};
+template <>
+struct handle_type_name<handle> {
+    static constexpr auto name = handle_type_name<object>::name;
+};
+template <>
 struct handle_type_name<none> {
     static constexpr auto name = const_name("None");
 };
 template <>
+struct handle_type_name<sequence> {
+    static constexpr auto name = const_name("Sequence");
+};
+template <>
+struct handle_type_name<bytearray> {
+    static constexpr auto name = const_name("bytearray");
+};
+template <>
+struct handle_type_name<memoryview> {
+    static constexpr auto name = const_name("memoryview");
+};
+template <>
+struct handle_type_name<slice> {
+    static constexpr auto name = const_name("slice");
+};
+template <>
+struct handle_type_name<type> {
+    static constexpr auto name = const_name("type");
+};
+template <>
+struct handle_type_name<capsule> {
+    static constexpr auto name = const_name("capsule");
+};
+template <>
+struct handle_type_name<ellipsis> {
+    static constexpr auto name = const_name("ellipsis");
+};
+template <>
+struct handle_type_name<weakref> {
+    static constexpr auto name = const_name("weakref");
+};
+template <>
 struct handle_type_name<args> {
     static constexpr auto name = const_name("*args");
 };
@@ -904,9 +1016,41 @@ template <>
 struct handle_type_name<kwargs> {
     static constexpr auto name = const_name("**kwargs");
 };
+template <>
+struct handle_type_name<obj_attr_accessor> {
+    static constexpr auto name = const_name<obj_attr_accessor>();
+};
+template <>
+struct handle_type_name<str_attr_accessor> {
+    static constexpr auto name = const_name<str_attr_accessor>();
+};
+template <>
+struct handle_type_name<item_accessor> {
+    static constexpr auto name = const_name<item_accessor>();
+};
+template <>
+struct handle_type_name<sequence_accessor> {
+    static constexpr auto name = const_name<sequence_accessor>();
+};
+template <>
+struct handle_type_name<list_accessor> {
+    static constexpr auto name = const_name<list_accessor>();
+};
+template <>
+struct handle_type_name<tuple_accessor> {
+    static constexpr auto name = const_name<tuple_accessor>();
+};
 
 template <typename type>
 struct pyobject_caster {
+    template <typename T = type, enable_if_t<std::is_same<T, handle>::value, int> = 0>
+    pyobject_caster() : value() {}
+
+    // `type` may not be default constructible (e.g. frozenset, anyset).  Initializing `value`
+    // to a nil handle is safe since it will only be accessed if `load` succeeds.
+    template <typename T = type, enable_if_t<std::is_base_of<object, T>::value, int> = 0>
+    pyobject_caster() : value(reinterpret_steal<type>(handle())) {}
+
     template <typename T = type, enable_if_t<std::is_same<T, handle>::value, int> = 0>
     bool load(handle src, bool /* convert */) {
         value = src;
@@ -951,7 +1095,7 @@ struct move_always<
     enable_if_t<
         all_of<move_is_plain_type<T>,
                negation<is_copy_constructible<T>>,
-               std::is_move_constructible<T>,
+               is_move_constructible<T>,
                std::is_same<decltype(std::declval<make_caster<T>>().operator T &()), T &>>::value>>
     : std::true_type {};
 template <typename T, typename SFINAE = void>
@@ -962,7 +1106,7 @@ struct move_if_unreferenced<
     enable_if_t<
         all_of<move_is_plain_type<T>,
                negation<move_always<T>>,
-               std::is_move_constructible<T>,
+               is_move_constructible<T>,
                std::is_same<decltype(std::declval<make_caster<T>>().operator T &()), T &>>::value>>
     : std::true_type {};
 template <typename T>
@@ -1000,13 +1144,18 @@ struct return_value_policy_override<
 // Basic python -> C++ casting; throws if casting fails
 template <typename T, typename SFINAE>
 type_caster<T, SFINAE> &load_type(type_caster<T, SFINAE> &conv, const handle &handle) {
+    static_assert(!detail::is_pyobject<T>::value,
+                  "Internal error: type_caster should only be used for C++ types");
     if (!conv.load(handle, true)) {
-#if defined(NDEBUG)
+#if !defined(PYBIND11_DETAILED_ERROR_MESSAGES)
         throw cast_error(
-            "Unable to cast Python instance to C++ type (compile in debug mode for details)");
+            "Unable to cast Python instance of type "
+            + str(type::handle_of(handle)).cast<std::string>()
+            + " to C++ type '?' (#define "
+              "PYBIND11_DETAILED_ERROR_MESSAGES or compile in debug mode for details)");
 #else
         throw cast_error("Unable to cast Python instance of type "
-                         + (std::string) str(type::handle_of(handle)) + " to C++ type '"
+                         + str(type::handle_of(handle)).cast<std::string>() + " to C++ type '"
                          + type_id<T>() + "'");
 #endif
     }
@@ -1023,7 +1172,11 @@ make_caster<T> load_type(const handle &handle) {
 PYBIND11_NAMESPACE_END(detail)
 
 // pytype -> C++ type
-template <typename T, detail::enable_if_t<!detail::is_pyobject<T>::value, int> = 0>
+template <typename T,
+          detail::enable_if_t<!detail::is_pyobject<T>::value
+                                  && !detail::is_same_ignoring_cvref<T, PyObject *>::value,
+                              int>
+          = 0>
 T cast(const handle &handle) {
     using namespace detail;
     static_assert(!cast_is_temporary_value_reference<T>::value,
@@ -1037,6 +1190,34 @@ T cast(const handle &handle) {
     return T(reinterpret_borrow<object>(handle));
 }
 
+// Note that `cast<PyObject *>(obj)` increments the reference count of `obj`.
+// This is necessary for the case that `obj` is a temporary, and could
+// not possibly be different, given
+// 1. the established convention that the passed `handle` is borrowed, and
+// 2. we don't want to force all generic code using `cast<T>()` to special-case
+//    handling of `T` = `PyObject *` (to increment the reference count there).
+// It is the responsibility of the caller to ensure that the reference count
+// is decremented.
+template <typename T,
+          typename Handle,
+          detail::enable_if_t<detail::is_same_ignoring_cvref<T, PyObject *>::value
+                                  && detail::is_same_ignoring_cvref<Handle, handle>::value,
+                              int>
+          = 0>
+T cast(Handle &&handle) {
+    return handle.inc_ref().ptr();
+}
+// To optimize way an inc_ref/dec_ref cycle:
+template <typename T,
+          typename Object,
+          detail::enable_if_t<detail::is_same_ignoring_cvref<T, PyObject *>::value
+                                  && detail::is_same_ignoring_cvref<Object, object>::value,
+                              int>
+          = 0>
+T cast(Object &&obj) {
+    return obj.release().ptr();
+}
+
 // C++ type -> py::object
 template <typename T, detail::enable_if_t<!detail::is_pyobject<T>::value, int> = 0>
 object cast(T &&value,
@@ -1068,14 +1249,15 @@ inline void handle::cast() const {
 template <typename T>
 detail::enable_if_t<!detail::move_never<T>::value, T> move(object &&obj) {
     if (obj.ref_count() > 1) {
-#if defined(NDEBUG)
+#if !defined(PYBIND11_DETAILED_ERROR_MESSAGES)
         throw cast_error(
-            "Unable to cast Python instance to C++ rvalue: instance has multiple references"
-            " (compile in debug mode for details)");
+            "Unable to cast Python " + str(type::handle_of(obj)).cast<std::string>()
+            + " instance to C++ rvalue: instance has multiple references"
+              " (#define PYBIND11_DETAILED_ERROR_MESSAGES or compile in debug mode for details)");
 #else
-        throw cast_error("Unable to move from Python " + (std::string) str(type::handle_of(obj))
-                         + " instance to C++ " + type_id<T>()
-                         + " instance: instance has multiple references");
+        throw cast_error("Unable to move from Python "
+                         + str(type::handle_of(obj)).cast<std::string>() + " instance to C++ "
+                         + type_id<T>() + " instance: instance has multiple references");
 #endif
     }
 
@@ -1090,21 +1272,30 @@ detail::enable_if_t<!detail::move_never<T>::value, T> move(object &&obj) {
 // - If both movable and copyable, check ref count: if 1, move; otherwise copy
 // - Otherwise (not movable), copy.
 template <typename T>
-detail::enable_if_t<detail::move_always<T>::value, T> cast(object &&object) {
+detail::enable_if_t<!detail::is_pyobject<T>::value && detail::move_always<T>::value, T>
+cast(object &&object) {
     return move<T>(std::move(object));
 }
 template <typename T>
-detail::enable_if_t<detail::move_if_unreferenced<T>::value, T> cast(object &&object) {
+detail::enable_if_t<!detail::is_pyobject<T>::value && detail::move_if_unreferenced<T>::value, T>
+cast(object &&object) {
     if (object.ref_count() > 1) {
         return cast<T>(object);
     }
     return move<T>(std::move(object));
 }
 template <typename T>
-detail::enable_if_t<detail::move_never<T>::value, T> cast(object &&object) {
+detail::enable_if_t<!detail::is_pyobject<T>::value && detail::move_never<T>::value, T>
+cast(object &&object) {
     return cast<T>(object);
 }
 
+// pytype rvalue -> pytype (calls converting constructor)
+template <typename T>
+detail::enable_if_t<detail::is_pyobject<T>::value, T> cast(object &&object) {
+    return T(std::move(object));
+}
+
 template <typename T>
 T object::cast() const & {
     return pybind11::cast<T>(*this);
@@ -1155,24 +1346,37 @@ enable_if_t<!cast_is_temporary_value_reference<T>::value, T> cast_ref(object &&,
 // static_assert, even though if it's in dead code, so we provide a "trampoline" to pybind11::cast
 // that only does anything in cases where pybind11::cast is valid.
 template <typename T>
-enable_if_t<!cast_is_temporary_value_reference<T>::value, T> cast_safe(object &&o) {
-    return pybind11::cast<T>(std::move(o));
+enable_if_t<cast_is_temporary_value_reference<T>::value
+                && !detail::is_same_ignoring_cvref<T, PyObject *>::value,
+            T>
+cast_safe(object &&) {
+    pybind11_fail("Internal error: cast_safe fallback invoked");
 }
 template <typename T>
-enable_if_t<cast_is_temporary_value_reference<T>::value, T> cast_safe(object &&) {
-    pybind11_fail("Internal error: cast_safe fallback invoked");
+enable_if_t<std::is_void<T>::value, void> cast_safe(object &&) {}
+template <typename T>
+enable_if_t<detail::is_same_ignoring_cvref<T, PyObject *>::value, PyObject *>
+cast_safe(object &&o) {
+    return o.release().ptr();
+}
+template <typename T>
+enable_if_t<detail::none_of<cast_is_temporary_value_reference<T>,
+                            detail::is_same_ignoring_cvref<T, PyObject *>,
+                            std::is_void<T>>::value,
+            T>
+cast_safe(object &&o) {
+    return pybind11::cast<T>(std::move(o));
 }
-template <>
-inline void cast_safe<void>(object &&) {}
 
 PYBIND11_NAMESPACE_END(detail)
 
 // The overloads could coexist, i.e. the #if is not strictly speaking needed,
 // but it is an easy minor optimization.
-#if defined(NDEBUG)
-inline cast_error cast_error_unable_to_convert_call_arg() {
-    return cast_error(
-        "Unable to convert call argument to Python object (compile in debug mode for details)");
+#if !defined(PYBIND11_DETAILED_ERROR_MESSAGES)
+inline cast_error cast_error_unable_to_convert_call_arg(const std::string &name) {
+    return cast_error("Unable to convert call argument '" + name
+                      + "' to Python object (#define "
+                        "PYBIND11_DETAILED_ERROR_MESSAGES or compile in debug mode for details)");
 }
 #else
 inline cast_error cast_error_unable_to_convert_call_arg(const std::string &name,
@@ -1194,8 +1398,8 @@ tuple make_tuple(Args &&...args_) {
         detail::make_caster<Args>::cast(std::forward<Args>(args_), policy, nullptr))...}};
     for (size_t i = 0; i < args.size(); i++) {
         if (!args[i]) {
-#if defined(NDEBUG)
-            throw cast_error_unable_to_convert_call_arg();
+#if !defined(PYBIND11_DETAILED_ERROR_MESSAGES)
+            throw cast_error_unable_to_convert_call_arg(std::to_string(i));
 #else
             std::array<std::string, size> argtypes{{type_id<Args>()...}};
             throw cast_error_unable_to_convert_call_arg(std::to_string(i), argtypes[i]);
@@ -1243,10 +1447,10 @@ struct arg_v : arg {
 private:
     template <typename T>
     arg_v(arg &&base, T &&x, const char *descr = nullptr)
-        : arg(base), value(reinterpret_steal<object>(
-                         detail::make_caster<T>::cast(x, return_value_policy::automatic, {}))),
+        : arg(base), value(reinterpret_steal<object>(detail::make_caster<T>::cast(
+                         std::forward<T>(x), return_value_policy::automatic, {}))),
           descr(descr)
-#if !defined(NDEBUG)
+#if defined(PYBIND11_DETAILED_ERROR_MESSAGES)
           ,
           type(type_id<T>())
 #endif
@@ -1286,7 +1490,7 @@ struct arg_v : arg {
     object value;
     /// The (optional) description of the default value
     const char *descr;
-#if !defined(NDEBUG)
+#if defined(PYBIND11_DETAILED_ERROR_MESSAGES)
     /// The C++ type name of the default value (only available when compiled in debug mode)
     std::string type;
 #endif
@@ -1315,7 +1519,15 @@ inline namespace literals {
 /** \rst
     String literal version of `arg`
  \endrst */
-constexpr arg operator"" _a(const char *name, size_t) { return arg(name); }
+constexpr arg
+#if !defined(__clang__) && defined(__GNUC__) && __GNUC__ < 5
+operator"" _a // gcc 4.8.5 insists on having a space (hard error).
+#else
+operator""_a // clang 17 generates a deprecation warning if there is a space.
+#endif
+    (const char *name, size_t) {
+    return arg(name);
+}
 } // namespace literals
 
 PYBIND11_NAMESPACE_BEGIN(detail)
@@ -1376,7 +1588,8 @@ class argument_loader {
     static_assert(args_pos == -1 || args_pos == constexpr_first<argument_is_args, Args...>(),
                   "py::args cannot be specified more than once");
 
-    static constexpr auto arg_names = concat(type_descr(make_caster<Args>::name)...);
+    static constexpr auto arg_names
+        = ::pybind11::detail::concat(type_descr(make_caster<Args>::name)...);
 
     bool load_args(function_call &call) { return load_impl_sequence(call, indices{}); }
 
@@ -1484,14 +1697,14 @@ class unpacking_collector {
         auto o = reinterpret_steal<object>(
             detail::make_caster<T>::cast(std::forward<T>(x), policy, {}));
         if (!o) {
-#if defined(NDEBUG)
-            throw cast_error_unable_to_convert_call_arg();
+#if !defined(PYBIND11_DETAILED_ERROR_MESSAGES)
+            throw cast_error_unable_to_convert_call_arg(std::to_string(args_list.size()));
 #else
             throw cast_error_unable_to_convert_call_arg(std::to_string(args_list.size()),
                                                         type_id<T>());
 #endif
         }
-        args_list.append(o);
+        args_list.append(std::move(o));
     }
 
     void process(list &args_list, detail::args_proxy ap) {
@@ -1502,27 +1715,27 @@ class unpacking_collector {
 
     void process(list & /*args_list*/, arg_v a) {
         if (!a.name) {
-#if defined(NDEBUG)
+#if !defined(PYBIND11_DETAILED_ERROR_MESSAGES)
             nameless_argument_error();
 #else
             nameless_argument_error(a.type);
 #endif
         }
         if (m_kwargs.contains(a.name)) {
-#if defined(NDEBUG)
+#if !defined(PYBIND11_DETAILED_ERROR_MESSAGES)
             multiple_values_error();
 #else
             multiple_values_error(a.name);
 #endif
         }
         if (!a.value) {
-#if defined(NDEBUG)
-            throw cast_error_unable_to_convert_call_arg();
+#if !defined(PYBIND11_DETAILED_ERROR_MESSAGES)
+            throw cast_error_unable_to_convert_call_arg(a.name);
 #else
             throw cast_error_unable_to_convert_call_arg(a.name, a.type);
 #endif
         }
-        m_kwargs[a.name] = a.value;
+        m_kwargs[a.name] = std::move(a.value);
     }
 
     void process(list & /*args_list*/, detail::kwargs_proxy kp) {
@@ -1531,7 +1744,7 @@ class unpacking_collector {
         }
         for (auto k : reinterpret_borrow<dict>(kp)) {
             if (m_kwargs.contains(k.first)) {
-#if defined(NDEBUG)
+#if !defined(PYBIND11_DETAILED_ERROR_MESSAGES)
                 multiple_values_error();
 #else
                 multiple_values_error(str(k.first));
@@ -1542,9 +1755,10 @@ class unpacking_collector {
     }
 
     [[noreturn]] static void nameless_argument_error() {
-        throw type_error("Got kwargs without a name; only named arguments "
-                         "may be passed via py::arg() to a python function call. "
-                         "(compile in debug mode for details)");
+        throw type_error(
+            "Got kwargs without a name; only named arguments "
+            "may be passed via py::arg() to a python function call. "
+            "(#define PYBIND11_DETAILED_ERROR_MESSAGES or compile in debug mode for details)");
     }
     [[noreturn]] static void nameless_argument_error(const std::string &type) {
         throw type_error("Got kwargs without a name of type '" + type
@@ -1552,8 +1766,9 @@ class unpacking_collector {
                            "arguments may be passed via py::arg() to a python function call. ");
     }
     [[noreturn]] static void multiple_values_error() {
-        throw type_error("Got multiple values for keyword argument "
-                         "(compile in debug mode for details)");
+        throw type_error(
+            "Got multiple values for keyword argument "
+            "(#define PYBIND11_DETAILED_ERROR_MESSAGES or compile in debug mode for details)");
     }
 
     [[noreturn]] static void multiple_values_error(const std::string &name) {
@@ -1625,12 +1840,12 @@ handle type::handle_of() {
 }
 
 #define PYBIND11_MAKE_OPAQUE(...)                                                                 \
-    namespace pybind11 {                                                                          \
+    PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)                                                  \
     namespace detail {                                                                            \
     template <>                                                                                   \
     class type_caster<__VA_ARGS__> : public type_caster_base<__VA_ARGS__> {};                     \
     }                                                                                             \
-    }
+    PYBIND11_NAMESPACE_END(PYBIND11_NAMESPACE)
 
 /// Lets you pass a type containing a `,` through a macro parameter without needing a separate
 /// typedef, e.g.:
diff --git a/include/pybind11/detail/class.h b/include/pybind11/detail/class.h
index e52ec1d3e6..d30621c88c 100644
--- a/include/pybind11/detail/class.h
+++ b/include/pybind11/detail/class.h
@@ -55,6 +55,9 @@ extern "C" inline int pybind11_static_set(PyObject *self, PyObject *obj, PyObjec
     return PyProperty_Type.tp_descr_set(self, cls, value);
 }
 
+// Forward declaration to use in `make_static_property_type()`
+inline void enable_dynamic_attributes(PyHeapTypeObject *heap_type);
+
 /** A `static_property` is the same as a `property` but the `__get__()` and `__set__()`
     methods are modified to always use the object type instead of a concrete instance.
     Return value: New reference. */
@@ -83,6 +86,12 @@ inline PyTypeObject *make_static_property_type() {
     type->tp_descr_get = pybind11_static_get;
     type->tp_descr_set = pybind11_static_set;
 
+#    if PY_VERSION_HEX >= 0x030C0000
+    // Since Python-3.12 property-derived types are required to
+    // have dynamic attributes (to set `__doc__`)
+    enable_dynamic_attributes(heap_type);
+#    endif
+
     if (PyType_Ready(type) < 0) {
         pybind11_fail("make_static_property_type(): failure in PyType_Ready()!");
     }
@@ -179,12 +188,10 @@ extern "C" inline PyObject *pybind11_meta_call(PyObject *type, PyObject *args, P
         return nullptr;
     }
 
-    // This must be a pybind11 instance
-    auto *instance = reinterpret_cast<detail::instance *>(self);
-
     // Ensure that the base __init__ function(s) were called
-    for (const auto &vh : values_and_holders(instance)) {
-        if (!vh.holder_constructed()) {
+    values_and_holders vhs(self);
+    for (const auto &vh : vhs) {
+        if (!vh.holder_constructed() && !vhs.is_redundant_value_and_holder(vh)) {
             PyErr_Format(PyExc_TypeError,
                          "%.200s.__init__() must be called when overriding __init__",
                          get_fully_qualified_tp_name(vh.type->type).c_str());
@@ -198,39 +205,40 @@ extern "C" inline PyObject *pybind11_meta_call(PyObject *type, PyObject *args, P
 
 /// Cleanup the type-info for a pybind11-registered type.
 extern "C" inline void pybind11_meta_dealloc(PyObject *obj) {
-    auto *type = (PyTypeObject *) obj;
-    auto &internals = get_internals();
-
-    // A pybind11-registered type will:
-    // 1) be found in internals.registered_types_py
-    // 2) have exactly one associated `detail::type_info`
-    auto found_type = internals.registered_types_py.find(type);
-    if (found_type != internals.registered_types_py.end() && found_type->second.size() == 1
-        && found_type->second[0]->type == type) {
-
-        auto *tinfo = found_type->second[0];
-        auto tindex = std::type_index(*tinfo->cpptype);
-        internals.direct_conversions.erase(tindex);
-
-        if (tinfo->module_local) {
-            get_local_internals().registered_types_cpp.erase(tindex);
-        } else {
-            internals.registered_types_cpp.erase(tindex);
-        }
-        internals.registered_types_py.erase(tinfo->type);
-
-        // Actually just `std::erase_if`, but that's only available in C++20
-        auto &cache = internals.inactive_override_cache;
-        for (auto it = cache.begin(), last = cache.end(); it != last;) {
-            if (it->first == (PyObject *) tinfo->type) {
-                it = cache.erase(it);
+    with_internals([obj](internals &internals) {
+        auto *type = (PyTypeObject *) obj;
+
+        // A pybind11-registered type will:
+        // 1) be found in internals.registered_types_py
+        // 2) have exactly one associated `detail::type_info`
+        auto found_type = internals.registered_types_py.find(type);
+        if (found_type != internals.registered_types_py.end() && found_type->second.size() == 1
+            && found_type->second[0]->type == type) {
+
+            auto *tinfo = found_type->second[0];
+            auto tindex = std::type_index(*tinfo->cpptype);
+            internals.direct_conversions.erase(tindex);
+
+            if (tinfo->module_local) {
+                get_local_internals().registered_types_cpp.erase(tindex);
             } else {
-                ++it;
+                internals.registered_types_cpp.erase(tindex);
+            }
+            internals.registered_types_py.erase(tinfo->type);
+
+            // Actually just `std::erase_if`, but that's only available in C++20
+            auto &cache = internals.inactive_override_cache;
+            for (auto it = cache.begin(), last = cache.end(); it != last;) {
+                if (it->first == (PyObject *) tinfo->type) {
+                    it = cache.erase(it);
+                } else {
+                    ++it;
+                }
             }
-        }
 
-        delete tinfo;
-    }
+            delete tinfo;
+        }
+    });
 
     PyType_Type.tp_dealloc(obj);
 }
@@ -303,19 +311,20 @@ inline void traverse_offset_bases(void *valueptr,
 }
 
 inline bool register_instance_impl(void *ptr, instance *self) {
-    get_internals().registered_instances.emplace(ptr, self);
+    with_instance_map(ptr, [&](instance_map &instances) { instances.emplace(ptr, self); });
     return true; // unused, but gives the same signature as the deregister func
 }
 inline bool deregister_instance_impl(void *ptr, instance *self) {
-    auto &registered_instances = get_internals().registered_instances;
-    auto range = registered_instances.equal_range(ptr);
-    for (auto it = range.first; it != range.second; ++it) {
-        if (self == it->second) {
-            registered_instances.erase(it);
-            return true;
+    return with_instance_map(ptr, [&](instance_map &instances) {
+        auto range = instances.equal_range(ptr);
+        for (auto it = range.first; it != range.second; ++it) {
+            if (self == it->second) {
+                instances.erase(it);
+                return true;
+            }
         }
-    }
-    return false;
+        return false;
+    });
 }
 
 inline void register_instance(instance *self, void *valptr, const type_info *tinfo) {
@@ -365,28 +374,37 @@ extern "C" inline PyObject *pybind11_object_new(PyTypeObject *type, PyObject *,
 extern "C" inline int pybind11_object_init(PyObject *self, PyObject *, PyObject *) {
     PyTypeObject *type = Py_TYPE(self);
     std::string msg = get_fully_qualified_tp_name(type) + ": No constructor defined!";
-    PyErr_SetString(PyExc_TypeError, msg.c_str());
+    set_error(PyExc_TypeError, msg.c_str());
     return -1;
 }
 
 inline void add_patient(PyObject *nurse, PyObject *patient) {
-    auto &internals = get_internals();
     auto *instance = reinterpret_cast<detail::instance *>(nurse);
     instance->has_patients = true;
     Py_INCREF(patient);
-    internals.patients[nurse].push_back(patient);
+
+    with_internals([&](internals &internals) { internals.patients[nurse].push_back(patient); });
 }
 
 inline void clear_patients(PyObject *self) {
     auto *instance = reinterpret_cast<detail::instance *>(self);
-    auto &internals = get_internals();
-    auto pos = internals.patients.find(self);
-    assert(pos != internals.patients.end());
-    // Clearing the patients can cause more Python code to run, which
-    // can invalidate the iterator. Extract the vector of patients
-    // from the unordered_map first.
-    auto patients = std::move(pos->second);
-    internals.patients.erase(pos);
+    std::vector<PyObject *> patients;
+
+    with_internals([&](internals &internals) {
+        auto pos = internals.patients.find(self);
+
+        if (pos == internals.patients.end()) {
+            pybind11_fail(
+                "FATAL: Internal consistency check failed: Invalid clear_patients() call.");
+        }
+
+        // Clearing the patients can cause more Python code to run, which
+        // can invalidate the iterator. Extract the vector of patients
+        // from the unordered_map first.
+        patients = std::move(pos->second);
+        internals.patients.erase(pos);
+    });
+
     instance->has_patients = false;
     for (PyObject *&patient : patients) {
         Py_CLEAR(patient);
@@ -435,9 +453,17 @@ inline void clear_instance(PyObject *self) {
 /// Instance destructor function for all pybind11 types. It calls `type_info.dealloc`
 /// to destroy the C++ object itself, while the rest is Python bookkeeping.
 extern "C" inline void pybind11_object_dealloc(PyObject *self) {
+    auto *type = Py_TYPE(self);
+
+    // If this is a GC tracked object, untrack it first
+    // Note that the track call is implicitly done by the
+    // default tp_alloc, which we never override.
+    if (PyType_HasFeature(type, Py_TPFLAGS_HAVE_GC) != 0) {
+        PyObject_GC_UnTrack(self);
+    }
+
     clear_instance(self);
 
-    auto *type = Py_TYPE(self);
     type->tp_free(self);
 
 #if PY_VERSION_HEX < 0x03080000
@@ -455,6 +481,8 @@ extern "C" inline void pybind11_object_dealloc(PyObject *self) {
 #endif
 }
 
+std::string error_string();
+
 /** Create the type which can be used as a common base for all classes.  This is
     needed in order to satisfy Python's requirements for multiple inheritance.
     Return value: New reference. */
@@ -490,7 +518,7 @@ inline PyObject *make_object_base_type(PyTypeObject *metaclass) {
     type->tp_weaklistoffset = offsetof(instance, weakrefs);
 
     if (PyType_Ready(type) < 0) {
-        pybind11_fail("PyType_Ready failed in make_object_base_type():" + error_string());
+        pybind11_fail("PyType_Ready failed in make_object_base_type(): " + error_string());
     }
 
     setattr((PyObject *) type, "__module__", str("pybind11_builtins"));
@@ -500,42 +528,29 @@ inline PyObject *make_object_base_type(PyTypeObject *metaclass) {
     return (PyObject *) heap_type;
 }
 
-/// dynamic_attr: Support for `d = instance.__dict__`.
-extern "C" inline PyObject *pybind11_get_dict(PyObject *self, void *) {
-    PyObject *&dict = *_PyObject_GetDictPtr(self);
-    if (!dict) {
-        dict = PyDict_New();
-    }
-    Py_XINCREF(dict);
-    return dict;
-}
-
-/// dynamic_attr: Support for `instance.__dict__ = dict()`.
-extern "C" inline int pybind11_set_dict(PyObject *self, PyObject *new_dict, void *) {
-    if (!PyDict_Check(new_dict)) {
-        PyErr_Format(PyExc_TypeError,
-                     "__dict__ must be set to a dictionary, not a '%.200s'",
-                     get_fully_qualified_tp_name(Py_TYPE(new_dict)).c_str());
-        return -1;
-    }
-    PyObject *&dict = *_PyObject_GetDictPtr(self);
-    Py_INCREF(new_dict);
-    Py_CLEAR(dict);
-    dict = new_dict;
-    return 0;
-}
-
 /// dynamic_attr: Allow the garbage collector to traverse the internal instance `__dict__`.
 extern "C" inline int pybind11_traverse(PyObject *self, visitproc visit, void *arg) {
+#if PY_VERSION_HEX >= 0x030D0000
+    PyObject_VisitManagedDict(self, visit, arg);
+#else
     PyObject *&dict = *_PyObject_GetDictPtr(self);
     Py_VISIT(dict);
+#endif
+// https://docs.python.org/3/c-api/typeobj.html#c.PyTypeObject.tp_traverse
+#if PY_VERSION_HEX >= 0x03090000
+    Py_VISIT(Py_TYPE(self));
+#endif
     return 0;
 }
 
 /// dynamic_attr: Allow the GC to clear the dictionary.
 extern "C" inline int pybind11_clear(PyObject *self) {
+#if PY_VERSION_HEX >= 0x030D0000
+    PyObject_ClearManagedDict(self);
+#else
     PyObject *&dict = *_PyObject_GetDictPtr(self);
     Py_CLEAR(dict);
+#endif
     return 0;
 }
 
@@ -543,14 +558,18 @@ extern "C" inline int pybind11_clear(PyObject *self) {
 inline void enable_dynamic_attributes(PyHeapTypeObject *heap_type) {
     auto *type = &heap_type->ht_type;
     type->tp_flags |= Py_TPFLAGS_HAVE_GC;
+#if PY_VERSION_HEX < 0x030B0000
     type->tp_dictoffset = type->tp_basicsize;           // place dict at the end
     type->tp_basicsize += (ssize_t) sizeof(PyObject *); // and allocate enough space for it
+#else
+    type->tp_flags |= Py_TPFLAGS_MANAGED_DICT;
+#endif
     type->tp_traverse = pybind11_traverse;
     type->tp_clear = pybind11_clear;
 
-    static PyGetSetDef getset[] = {
-        {const_cast<char *>("__dict__"), pybind11_get_dict, pybind11_set_dict, nullptr, nullptr},
-        {nullptr, nullptr, nullptr, nullptr, nullptr}};
+    static PyGetSetDef getset[]
+        = {{"__dict__", PyObject_GenericGetDict, PyObject_GenericSetDict, nullptr, nullptr},
+           {nullptr, nullptr, nullptr, nullptr, nullptr}};
     type->tp_getset = getset;
 }
 
@@ -568,7 +587,7 @@ extern "C" inline int pybind11_getbuffer(PyObject *obj, Py_buffer *view, int fla
         if (view) {
             view->obj = nullptr;
         }
-        PyErr_SetString(PyExc_BufferError, "pybind11_getbuffer(): Internal error");
+        set_error(PyExc_BufferError, "pybind11_getbuffer(): Internal error");
         return -1;
     }
     std::memset(view, 0, sizeof(Py_buffer));
@@ -576,7 +595,7 @@ extern "C" inline int pybind11_getbuffer(PyObject *obj, Py_buffer *view, int fla
     if ((flags & PyBUF_WRITABLE) == PyBUF_WRITABLE && info->readonly) {
         delete info;
         // view->obj = nullptr;  // Was just memset to 0, so not necessary
-        PyErr_SetString(PyExc_BufferError, "Writable buffer requested for readonly storage");
+        set_error(PyExc_BufferError, "Writable buffer requested for readonly storage");
         return -1;
     }
     view->obj = obj;
@@ -642,10 +661,13 @@ inline PyObject *make_new_python_type(const type_record &rec) {
 
     char *tp_doc = nullptr;
     if (rec.doc && options::show_user_defined_docstrings()) {
-        /* Allocate memory for docstring (using PyObject_MALLOC, since
-           Python will free this later on) */
+        /* Allocate memory for docstring (Python will free this later on) */
         size_t size = std::strlen(rec.doc) + 1;
+#if PY_VERSION_HEX >= 0x030D0000
+        tp_doc = (char *) PyMem_MALLOC(size);
+#else
         tp_doc = (char *) PyObject_MALLOC(size);
+#endif
         std::memcpy((void *) tp_doc, rec.doc, size);
     }
 
@@ -707,7 +729,7 @@ inline PyObject *make_new_python_type(const type_record &rec) {
     }
 
     if (PyType_Ready(type) < 0) {
-        pybind11_fail(std::string(rec.name) + ": PyType_Ready failed (" + error_string() + ")!");
+        pybind11_fail(std::string(rec.name) + ": PyType_Ready failed: " + error_string());
     }
 
     assert(!rec.dynamic_attr || PyType_HasFeature(type, Py_TPFLAGS_HAVE_GC));
diff --git a/include/pybind11/detail/common.h b/include/pybind11/detail/common.h
index ea09bb3fdf..fc66e6f756 100644
--- a/include/pybind11/detail/common.h
+++ b/include/pybind11/detail/common.h
@@ -10,15 +10,76 @@
 #pragma once
 
 #define PYBIND11_VERSION_MAJOR 2
-#define PYBIND11_VERSION_MINOR 10
+#define PYBIND11_VERSION_MINOR 14
 #define PYBIND11_VERSION_PATCH 0.dev1
 
 // Similar to Python's convention: https://docs.python.org/3/c-api/apiabiversion.html
 // Additional convention: 0xD = dev
-#define PYBIND11_VERSION_HEX 0x020A00D1
+#define PYBIND11_VERSION_HEX 0x020E00D1
+
+// Define some generic pybind11 helper macros for warning management.
+//
+// Note that compiler-specific push/pop pairs are baked into the
+// PYBIND11_NAMESPACE_BEGIN/PYBIND11_NAMESPACE_END pair of macros. Therefore manual
+// PYBIND11_WARNING_PUSH/PYBIND11_WARNING_POP are usually only needed in `#include` sections.
+//
+// If you find you need to suppress a warning, please try to make the suppression as local as
+// possible using these macros. Please also be sure to push/pop with the pybind11 macros. Please
+// only use compiler specifics if you need to check specific versions, e.g. Apple Clang vs. vanilla
+// Clang.
+#if defined(_MSC_VER)
+#    define PYBIND11_COMPILER_MSVC
+#    define PYBIND11_PRAGMA(...) __pragma(__VA_ARGS__)
+#    define PYBIND11_WARNING_PUSH PYBIND11_PRAGMA(warning(push))
+#    define PYBIND11_WARNING_POP PYBIND11_PRAGMA(warning(pop))
+#elif defined(__INTEL_COMPILER)
+#    define PYBIND11_COMPILER_INTEL
+#    define PYBIND11_PRAGMA(...) _Pragma(#__VA_ARGS__)
+#    define PYBIND11_WARNING_PUSH PYBIND11_PRAGMA(warning push)
+#    define PYBIND11_WARNING_POP PYBIND11_PRAGMA(warning pop)
+#elif defined(__clang__)
+#    define PYBIND11_COMPILER_CLANG
+#    define PYBIND11_PRAGMA(...) _Pragma(#__VA_ARGS__)
+#    define PYBIND11_WARNING_PUSH PYBIND11_PRAGMA(clang diagnostic push)
+#    define PYBIND11_WARNING_POP PYBIND11_PRAGMA(clang diagnostic push)
+#elif defined(__GNUC__)
+#    define PYBIND11_COMPILER_GCC
+#    define PYBIND11_PRAGMA(...) _Pragma(#__VA_ARGS__)
+#    define PYBIND11_WARNING_PUSH PYBIND11_PRAGMA(GCC diagnostic push)
+#    define PYBIND11_WARNING_POP PYBIND11_PRAGMA(GCC diagnostic pop)
+#endif
+
+#ifdef PYBIND11_COMPILER_MSVC
+#    define PYBIND11_WARNING_DISABLE_MSVC(name) PYBIND11_PRAGMA(warning(disable : name))
+#else
+#    define PYBIND11_WARNING_DISABLE_MSVC(name)
+#endif
+
+#ifdef PYBIND11_COMPILER_CLANG
+#    define PYBIND11_WARNING_DISABLE_CLANG(name) PYBIND11_PRAGMA(clang diagnostic ignored name)
+#else
+#    define PYBIND11_WARNING_DISABLE_CLANG(name)
+#endif
+
+#ifdef PYBIND11_COMPILER_GCC
+#    define PYBIND11_WARNING_DISABLE_GCC(name) PYBIND11_PRAGMA(GCC diagnostic ignored name)
+#else
+#    define PYBIND11_WARNING_DISABLE_GCC(name)
+#endif
+
+#ifdef PYBIND11_COMPILER_INTEL
+#    define PYBIND11_WARNING_DISABLE_INTEL(name) PYBIND11_PRAGMA(warning disable name)
+#else
+#    define PYBIND11_WARNING_DISABLE_INTEL(name)
+#endif
+
+#define PYBIND11_NAMESPACE_BEGIN(name)                                                            \
+    namespace name {                                                                              \
+    PYBIND11_WARNING_PUSH
 
-#define PYBIND11_NAMESPACE_BEGIN(name) namespace name {
-#define PYBIND11_NAMESPACE_END(name) }
+#define PYBIND11_NAMESPACE_END(name)                                                              \
+    PYBIND11_WARNING_POP                                                                          \
+    }
 
 // Robust support for some features and loading modules compiled against different pybind versions
 // requires forcing hidden visibility on pybind code, so we enforce this by setting the attribute
@@ -38,6 +99,7 @@
 #            define PYBIND11_CPP17
 #            if __cplusplus >= 202002L
 #                define PYBIND11_CPP20
+// Please update tests/pybind11_tests.cpp `cpp_std()` when adding a macro here.
 #            endif
 #        endif
 #    endif
@@ -56,6 +118,14 @@
 #    endif
 #endif
 
+#if defined(PYBIND11_CPP20)
+#    define PYBIND11_CONSTINIT constinit
+#    define PYBIND11_DTOR_CONSTEXPR constexpr
+#else
+#    define PYBIND11_CONSTINIT
+#    define PYBIND11_DTOR_CONSTEXPR
+#endif
+
 // Compiler version assertions
 #if defined(__INTEL_COMPILER)
 #    if __INTEL_COMPILER < 1800
@@ -95,13 +165,10 @@
 #endif
 
 #if !defined(PYBIND11_EXPORT_EXCEPTION)
-#    ifdef __MINGW32__
-// workaround for:
-// error: 'dllexport' implies default visibility, but xxx has already been declared with a
-// different visibility
-#        define PYBIND11_EXPORT_EXCEPTION
-#    else
+#    if defined(__apple_build_version__)
 #        define PYBIND11_EXPORT_EXCEPTION PYBIND11_EXPORT
+#    else
+#        define PYBIND11_EXPORT_EXCEPTION
 #    endif
 #endif
 
@@ -153,9 +220,9 @@
 
 /// Include Python header, disable linking to pythonX_d.lib on Windows in debug mode
 #if defined(_MSC_VER)
-#    pragma warning(push)
+PYBIND11_WARNING_PUSH
+PYBIND11_WARNING_DISABLE_MSVC(4505)
 // C4505: 'PySlice_GetIndicesEx': unreferenced local function has been removed (PyPy only)
-#    pragma warning(disable : 4505)
 #    if defined(_DEBUG) && !defined(Py_DEBUG)
 // Workaround for a VS 2022 issue.
 // NOTE: This workaround knowingly violates the Python.h include order requirement:
@@ -204,13 +271,9 @@
 #    endif
 #endif
 
-#if defined(__cpp_lib_char8_t) && __cpp_lib_char8_t >= 201811L
-#    define PYBIND11_HAS_U8STRING
-#endif
-
 #include <Python.h>
-#if PY_VERSION_HEX < 0x03060000
-#    error "PYTHON < 3.6 IS UNSUPPORTED. pybind11 v2.9 was the last to support Python 2 and 3.5."
+#if PY_VERSION_HEX < 0x03070000
+#    error "PYTHON < 3.7 IS UNSUPPORTED. pybind11 v2.12 was the last to support Python 3.6."
 #endif
 #include <frameobject.h>
 #include <pythread.h>
@@ -232,12 +295,20 @@
 #    undef copysign
 #endif
 
+#if defined(PYBIND11_NUMPY_1_ONLY)
+#    define PYBIND11_INTERNAL_NUMPY_1_ONLY_DETECTED
+#endif
+
+#if defined(PYPY_VERSION) && !defined(PYBIND11_SIMPLE_GIL_MANAGEMENT)
+#    define PYBIND11_SIMPLE_GIL_MANAGEMENT
+#endif
+
 #if defined(_MSC_VER)
 #    if defined(PYBIND11_DEBUG_MARKER)
 #        define _DEBUG
 #        undef PYBIND11_DEBUG_MARKER
 #    endif
-#    pragma warning(pop)
+PYBIND11_WARNING_POP
 #endif
 
 #include <cstddef>
@@ -258,6 +329,17 @@
 #    endif
 #endif
 
+// Must be after including <version> or one of the other headers specified by the standard
+#if defined(__cpp_lib_char8_t) && __cpp_lib_char8_t >= 201811L
+#    define PYBIND11_HAS_U8STRING
+#endif
+
+// See description of PR #4246:
+#if !defined(PYBIND11_NO_ASSERT_GIL_HELD_INCREF_DECREF) && !defined(NDEBUG)                       \
+    && !defined(PYPY_VERSION) && !defined(PYBIND11_ASSERT_GIL_HELD_INCREF_DECREF)
+#    define PYBIND11_ASSERT_GIL_HELD_INCREF_DECREF
+#endif
+
 // #define PYBIND11_STR_LEGACY_PERMISSIVE
 // If DEFINED, pybind11::str can hold PyUnicodeObject or PyBytesObject
 //             (probably surprising and never documented, but this was the
@@ -328,7 +410,7 @@
         return nullptr;                                                                           \
     }                                                                                             \
     catch (const std::exception &e) {                                                             \
-        PyErr_SetString(PyExc_ImportError, e.what());                                             \
+        ::pybind11::set_error(PyExc_ImportError, e.what());                                       \
         return nullptr;                                                                           \
     }
 
@@ -362,7 +444,7 @@
 
 /** \rst
     This macro creates the entry point that will be invoked when the Python interpreter
-    imports an extension module. The module name is given as the fist argument and it
+    imports an extension module. The module name is given as the first argument and it
     should not be in quotes. The second macro argument defines a variable of type
     `py::module_` which can be used to initialize the module.
 
@@ -381,7 +463,7 @@
             });
         }
 \endrst */
-#define PYBIND11_MODULE(name, variable)                                                           \
+#define PYBIND11_MODULE(name, variable, ...)                                                      \
     static ::pybind11::module_::module_def PYBIND11_CONCAT(pybind11_module_def_, name)            \
         PYBIND11_MAYBE_UNUSED;                                                                    \
     PYBIND11_MAYBE_UNUSED                                                                         \
@@ -390,7 +472,10 @@
         PYBIND11_CHECK_PYTHON_VERSION                                                             \
         PYBIND11_ENSURE_INTERNALS_READY                                                           \
         auto m = ::pybind11::module_::create_extension_module(                                    \
-            PYBIND11_TOSTRING(name), nullptr, &PYBIND11_CONCAT(pybind11_module_def_, name));      \
+            PYBIND11_TOSTRING(name),                                                              \
+            nullptr,                                                                              \
+            &PYBIND11_CONCAT(pybind11_module_def_, name),                                         \
+            ##__VA_ARGS__);                                                                       \
         try {                                                                                     \
             PYBIND11_CONCAT(pybind11_init_, name)(m);                                             \
             return m.ptr();                                                                       \
@@ -587,6 +672,10 @@ template <class T>
 using remove_cvref_t = typename remove_cvref<T>::type;
 #endif
 
+/// Example usage: is_same_ignoring_cvref<T, PyObject *>::value
+template <typename T, typename U>
+using is_same_ignoring_cvref = std::is_same<detail::remove_cvref_t<T>, U>;
+
 /// Index sequences
 #if defined(PYBIND11_CPP14)
 using std::index_sequence;
@@ -680,7 +769,16 @@ template <typename C, typename R, typename... A>
 struct remove_class<R (C::*)(A...) const> {
     using type = R(A...);
 };
-
+#ifdef __cpp_noexcept_function_type
+template <typename C, typename R, typename... A>
+struct remove_class<R (C::*)(A...) noexcept> {
+    using type = R(A...);
+};
+template <typename C, typename R, typename... A>
+struct remove_class<R (C::*)(A...) const noexcept> {
+    using type = R(A...);
+};
+#endif
 /// Helper template to strip away type modifiers
 template <typename T>
 struct intrinsic_type {
@@ -829,8 +927,7 @@ using is_template_base_of
     = decltype(is_template_base_of_impl<Base>::check((intrinsic_t<T> *) nullptr));
 #else
 struct is_template_base_of
-    : decltype(is_template_base_of_impl<Base>::check((intrinsic_t<T> *) nullptr)) {
-};
+    : decltype(is_template_base_of_impl<Base>::check((intrinsic_t<T> *) nullptr)){};
 #endif
 
 /// Check if T is an instantiation of the template `Class`. For example:
@@ -897,12 +994,6 @@ using expand_side_effects = bool[];
 
 PYBIND11_NAMESPACE_END(detail)
 
-#if defined(_MSC_VER)
-#    pragma warning(push)
-#    pragma warning(disable : 4275)
-//     warning C4275: An exported class was derived from a class that wasn't exported.
-//     Can be ignored when derived from a STL class.
-#endif
 /// C++ bindings of builtin Python exceptions
 class PYBIND11_EXPORT_EXCEPTION builtin_exception : public std::runtime_error {
 public:
@@ -910,9 +1001,6 @@ class PYBIND11_EXPORT_EXCEPTION builtin_exception : public std::runtime_error {
     /// Set the error using the Python C API
     virtual void set_error() const = 0;
 };
-#if defined(_MSC_VER)
-#    pragma warning(pop)
-#endif
 
 #define PYBIND11_RUNTIME_EXCEPTION(name, type)                                                    \
     class PYBIND11_EXPORT_EXCEPTION name : public builtin_exception {                             \
@@ -947,6 +1035,15 @@ PYBIND11_RUNTIME_EXCEPTION(reference_cast_error, PyExc_RuntimeError) /// Used in
 template <typename T, typename SFINAE = void>
 struct format_descriptor {};
 
+template <typename T>
+struct format_descriptor<
+    T,
+    detail::enable_if_t<detail::is_same_ignoring_cvref<T, PyObject *>::value>> {
+    static constexpr const char c = 'O';
+    static constexpr const char value[2] = {c, '\0'};
+    static std::string format() { return std::string(1, c); }
+};
+
 PYBIND11_NAMESPACE_BEGIN(detail)
 // Returns the index of the given type in the type char array below, and in the list in numpy.h
 // The order here is: bool; 8 ints ((signed,unsigned)x(8,16,32,64)bits); float,double,long double;
@@ -992,6 +1089,8 @@ constexpr const char
 struct error_scope {
     PyObject *type, *value, *trace;
     error_scope() { PyErr_Fetch(&type, &value, &trace); }
+    error_scope(const error_scope &) = delete;
+    error_scope &operator=(const error_scope &) = delete;
     ~error_scope() { PyErr_Restore(type, value, trace); }
 };
 
@@ -1010,14 +1109,14 @@ struct overload_cast_impl {
     }
 
     template <typename Return, typename Class>
-    constexpr auto operator()(Return (Class::*pmf)(Args...), std::false_type = {}) const noexcept
-        -> decltype(pmf) {
+    constexpr auto operator()(Return (Class::*pmf)(Args...),
+                              std::false_type = {}) const noexcept -> decltype(pmf) {
         return pmf;
     }
 
     template <typename Return, typename Class>
-    constexpr auto operator()(Return (Class::*pmf)(Args...) const, std::true_type) const noexcept
-        -> decltype(pmf) {
+    constexpr auto operator()(Return (Class::*pmf)(Args...) const,
+                              std::true_type) const noexcept -> decltype(pmf) {
         return pmf;
     }
 };
@@ -1030,12 +1129,7 @@ PYBIND11_NAMESPACE_END(detail)
 ///  - regular: static_cast<Return (Class::*)(Arg0, Arg1, Arg2)>(&Class::func)
 ///  - sweet:   overload_cast<Arg0, Arg1, Arg2>(&Class::func)
 template <typename... Args>
-#    if (defined(_MSC_VER) && _MSC_VER < 1920) /* MSVC 2017 */                                    \
-        || (defined(__clang__) && __clang_major__ == 5)
-static constexpr detail::overload_cast_impl<Args...> overload_cast = {};
-#    else
-static constexpr detail::overload_cast_impl<Args...> overload_cast;
-#    endif
+static constexpr detail::overload_cast_impl<Args...> overload_cast{};
 #endif
 
 /// Const member function selector for overload_cast
@@ -1144,15 +1238,30 @@ constexpr
 #    define PYBIND11_WORKAROUND_INCORRECT_GCC_UNUSED_BUT_SET_PARAMETER(...)
 #endif
 
-#if defined(_MSC_VER) // All versions (as of July 2021).
-
-// warning C4127: Conditional expression is constant
-constexpr inline bool silence_msvc_c4127(bool cond) { return cond; }
-
-#    define PYBIND11_SILENCE_MSVC_C4127(...) ::pybind11::detail::silence_msvc_c4127(__VA_ARGS__)
+#if defined(__clang__)                                                                            \
+    && (defined(__apple_build_version__) /* AppleClang 13.0.0.13000029 was the only data point    \
+                                            available. */                                         \
+        || (__clang_major__ >= 7                                                                  \
+            && __clang_major__ <= 12) /* Clang 3, 5, 13, 14, 15 do not generate the warning. */   \
+    )
+#    define PYBIND11_DETECTED_CLANG_WITH_MISLEADING_CALL_STD_MOVE_EXPLICITLY_WARNING
+// Example:
+// tests/test_kwargs_and_defaults.cpp:46:68: error: local variable 'args' will be copied despite
+// being returned by name [-Werror,-Wreturn-std-move]
+//     m.def("args_function", [](py::args args) -> py::tuple { return args; });
+//                                                                    ^~~~
+// test_kwargs_and_defaults.cpp:46:68: note: call 'std::move' explicitly to avoid copying
+//     m.def("args_function", [](py::args args) -> py::tuple { return args; });
+//                                                                    ^~~~
+//                                                                    std::move(args)
+#endif
 
-#else
-#    define PYBIND11_SILENCE_MSVC_C4127(...) __VA_ARGS__
+// Pybind offers detailed error messages by default for all builts that are debug (through the
+// negation of NDEBUG). This can also be manually enabled by users, for any builds, through
+// defining PYBIND11_DETAILED_ERROR_MESSAGES. This information is primarily useful for those
+// who are writing (as opposed to merely using) libraries that use pybind11.
+#if !defined(PYBIND11_DETAILED_ERROR_MESSAGES) && !defined(NDEBUG)
+#    define PYBIND11_DETAILED_ERROR_MESSAGES
 #endif
 
 PYBIND11_NAMESPACE_END(detail)
diff --git a/include/pybind11/detail/descr.h b/include/pybind11/detail/descr.h
index e7a5e2c145..7d546311e7 100644
--- a/include/pybind11/detail/descr.h
+++ b/include/pybind11/detail/descr.h
@@ -143,11 +143,25 @@ constexpr descr<N, Ts...> concat(const descr<N, Ts...> &descr) {
     return descr;
 }
 
+#ifdef __cpp_fold_expressions
+template <size_t N1, size_t N2, typename... Ts1, typename... Ts2>
+constexpr descr<N1 + N2 + 2, Ts1..., Ts2...> operator,(const descr<N1, Ts1...> &a,
+                                                       const descr<N2, Ts2...> &b) {
+    return a + const_name(", ") + b;
+}
+
 template <size_t N, typename... Ts, typename... Args>
-constexpr auto concat(const descr<N, Ts...> &d, const Args &...args)
-    -> decltype(std::declval<descr<N + 2, Ts...>>() + concat(args...)) {
+constexpr auto concat(const descr<N, Ts...> &d, const Args &...args) {
+    return (d, ..., args);
+}
+#else
+template <size_t N, typename... Ts, typename... Args>
+constexpr auto concat(const descr<N, Ts...> &d,
+                      const Args &...args) -> decltype(std::declval<descr<N + 2, Ts...>>()
+                                                       + concat(args...)) {
     return d + const_name(", ") + concat(args...);
 }
+#endif
 
 template <size_t N, typename... Ts>
 constexpr descr<N + 2, Ts...> type_descr(const descr<N, Ts...> &descr) {
diff --git a/include/pybind11/detail/init.h b/include/pybind11/detail/init.h
index e1e665a69b..4509bd131e 100644
--- a/include/pybind11/detail/init.h
+++ b/include/pybind11/detail/init.h
@@ -12,6 +12,9 @@
 #include "class.h"
 
 PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)
+
+PYBIND11_WARNING_DISABLE_MSVC(4127)
+
 PYBIND11_NAMESPACE_BEGIN(detail)
 
 template <>
@@ -62,7 +65,7 @@ constexpr bool is_alias(void *) {
 }
 
 // Constructs and returns a new object; if the given arguments don't map to a constructor, we fall
-// back to brace aggregate initiailization so that for aggregate initialization can be used with
+// back to brace aggregate initialization so that for aggregate initialization can be used with
 // py::init, e.g.  `py::init<int, int>` to initialize a `struct T { int a; int b; }`.  For
 // non-aggregate types, we need to use an ordinary T(...) constructor (invoking as `T{...}` usually
 // works, but will not do the expected thing when `T` has an `initializer_list<T>` constructor).
@@ -115,7 +118,7 @@ template <typename Class>
 void construct(value_and_holder &v_h, Cpp<Class> *ptr, bool need_alias) {
     PYBIND11_WORKAROUND_INCORRECT_MSVC_C4100(need_alias);
     no_nullptr(ptr);
-    if (PYBIND11_SILENCE_MSVC_C4127(Class::has_alias) && need_alias && !is_alias<Class>(ptr)) {
+    if (Class::has_alias && need_alias && !is_alias<Class>(ptr)) {
         // We're going to try to construct an alias by moving the cpp type.  Whether or not
         // that succeeds, we still need to destroy the original cpp pointer (either the
         // moved away leftover, if the alias construction works, or the value itself if we
@@ -156,7 +159,7 @@ void construct(value_and_holder &v_h, Holder<Class> holder, bool need_alias) {
     auto *ptr = holder_helper<Holder<Class>>::get(holder);
     no_nullptr(ptr);
     // If we need an alias, check that the held pointer is actually an alias instance
-    if (PYBIND11_SILENCE_MSVC_C4127(Class::has_alias) && need_alias && !is_alias<Class>(ptr)) {
+    if (Class::has_alias && need_alias && !is_alias<Class>(ptr)) {
         throw type_error("pybind11::init(): construction failed: returned holder-wrapped instance "
                          "is not an alias instance");
     }
@@ -172,9 +175,9 @@ void construct(value_and_holder &v_h, Holder<Class> holder, bool need_alias) {
 template <typename Class>
 void construct(value_and_holder &v_h, Cpp<Class> &&result, bool need_alias) {
     PYBIND11_WORKAROUND_INCORRECT_MSVC_C4100(need_alias);
-    static_assert(std::is_move_constructible<Cpp<Class>>::value,
+    static_assert(is_move_constructible<Cpp<Class>>::value,
                   "pybind11::init() return-by-value factory function requires a movable class");
-    if (PYBIND11_SILENCE_MSVC_C4127(Class::has_alias) && need_alias) {
+    if (Class::has_alias && need_alias) {
         construct_alias_from_cpp<Class>(is_alias_constructible<Class>{}, v_h, std::move(result));
     } else {
         v_h.value_ptr() = new Cpp<Class>(std::move(result));
@@ -187,7 +190,7 @@ void construct(value_and_holder &v_h, Cpp<Class> &&result, bool need_alias) {
 template <typename Class>
 void construct(value_and_holder &v_h, Alias<Class> &&result, bool) {
     static_assert(
-        std::is_move_constructible<Alias<Class>>::value,
+        is_move_constructible<Alias<Class>>::value,
         "pybind11::init() return-by-alias-value factory function requires a movable alias class");
     v_h.value_ptr() = new Alias<Class>(std::move(result));
 }
@@ -206,10 +209,11 @@ struct constructor {
             extra...);
     }
 
-    template <typename Class,
-              typename... Extra,
-              enable_if_t<Class::has_alias && std::is_constructible<Cpp<Class>, Args...>::value,
-                          int> = 0>
+    template <
+        typename Class,
+        typename... Extra,
+        enable_if_t<Class::has_alias && std::is_constructible<Cpp<Class>, Args...>::value, int>
+        = 0>
     static void execute(Class &cl, const Extra &...extra) {
         cl.def(
             "__init__",
@@ -226,10 +230,11 @@ struct constructor {
             extra...);
     }
 
-    template <typename Class,
-              typename... Extra,
-              enable_if_t<Class::has_alias && !std::is_constructible<Cpp<Class>, Args...>::value,
-                          int> = 0>
+    template <
+        typename Class,
+        typename... Extra,
+        enable_if_t<Class::has_alias && !std::is_constructible<Cpp<Class>, Args...>::value, int>
+        = 0>
     static void execute(Class &cl, const Extra &...extra) {
         cl.def(
             "__init__",
@@ -245,10 +250,11 @@ struct constructor {
 // Implementing class for py::init_alias<...>()
 template <typename... Args>
 struct alias_constructor {
-    template <typename Class,
-              typename... Extra,
-              enable_if_t<Class::has_alias && std::is_constructible<Alias<Class>, Args...>::value,
-                          int> = 0>
+    template <
+        typename Class,
+        typename... Extra,
+        enable_if_t<Class::has_alias && std::is_constructible<Alias<Class>, Args...>::value, int>
+        = 0>
     static void execute(Class &cl, const Extra &...extra) {
         cl.def(
             "__init__",
@@ -425,4 +431,4 @@ struct pickle_factory<Get, Set, RetState(Self), NewInstance(ArgState)> {
 
 PYBIND11_NAMESPACE_END(initimpl)
 PYBIND11_NAMESPACE_END(detail)
-PYBIND11_NAMESPACE_END(pybind11)
+PYBIND11_NAMESPACE_END(PYBIND11_NAMESPACE)
diff --git a/include/pybind11/detail/internals.h b/include/pybind11/detail/internals.h
index 3f83113bd7..3f0a6a9492 100644
--- a/include/pybind11/detail/internals.h
+++ b/include/pybind11/detail/internals.h
@@ -9,9 +9,17 @@
 
 #pragma once
 
+#include "common.h"
+
+#if defined(PYBIND11_SIMPLE_GIL_MANAGEMENT)
+#    include "../gil.h"
+#endif
+
 #include "../pytypes.h"
 
 #include <exception>
+#include <mutex>
+#include <thread>
 
 /// Tracks the `internals` and `type_info` ABI version independent of the main library version.
 ///
@@ -28,15 +36,27 @@
 /// further ABI-incompatible changes may be made before the ABI is officially
 /// changed to the new version.
 #ifndef PYBIND11_INTERNALS_VERSION
-#    define PYBIND11_INTERNALS_VERSION 4
+#    if PY_VERSION_HEX >= 0x030C0000 || defined(_MSC_VER)
+// Version bump for Python 3.12+, before first 3.12 beta release.
+// Version bump for MSVC piggy-backed on PR #4779. See comments there.
+#        define PYBIND11_INTERNALS_VERSION 5
+#    else
+#        define PYBIND11_INTERNALS_VERSION 4
+#    endif
 #endif
 
+// This requirement is mainly to reduce the support burden (see PR #4570).
+static_assert(PY_VERSION_HEX < 0x030C0000 || PYBIND11_INTERNALS_VERSION >= 5,
+              "pybind11 ABI version 5 is the minimum for Python 3.12+");
+
 PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)
 
 using ExceptionTranslator = void (*)(std::exception_ptr);
 
 PYBIND11_NAMESPACE_BEGIN(detail)
 
+constexpr const char *internals_function_record_capsule_name = "pybind11_function_record_capsule";
+
 // Forward declarations
 inline PyTypeObject *make_static_property_type();
 inline PyTypeObject *make_default_metaclass();
@@ -44,60 +64,41 @@ inline PyObject *make_object_base_type(PyTypeObject *metaclass);
 
 // The old Python Thread Local Storage (TLS) API is deprecated in Python 3.7 in favor of the new
 // Thread Specific Storage (TSS) API.
-#if PY_VERSION_HEX >= 0x03070000
 // Avoid unnecessary allocation of `Py_tss_t`, since we cannot use
 // `Py_LIMITED_API` anyway.
-#    if PYBIND11_INTERNALS_VERSION > 4
-#        define PYBIND11_TLS_KEY_REF Py_tss_t &
-#        ifdef __GNUC__
-// Clang on macOS warns due to `Py_tss_NEEDS_INIT` not specifying an initializer
-// for every field.
-#            define PYBIND11_TLS_KEY_INIT(var)                                                    \
-                _Pragma("GCC diagnostic push")                                         /**/       \
-                    _Pragma("GCC diagnostic ignored \"-Wmissing-field-initializers\"") /**/       \
-                    Py_tss_t var                                                                  \
-                    = Py_tss_NEEDS_INIT;                                                          \
-                _Pragma("GCC diagnostic pop")
-#        else
-#            define PYBIND11_TLS_KEY_INIT(var) Py_tss_t var = Py_tss_NEEDS_INIT;
-#        endif
-#        define PYBIND11_TLS_KEY_CREATE(var) (PyThread_tss_create(&(var)) == 0)
-#        define PYBIND11_TLS_GET_VALUE(key) PyThread_tss_get(&(key))
-#        define PYBIND11_TLS_REPLACE_VALUE(key, value) PyThread_tss_set(&(key), (value))
-#        define PYBIND11_TLS_DELETE_VALUE(key) PyThread_tss_set(&(key), nullptr)
-#        define PYBIND11_TLS_FREE(key) PyThread_tss_delete(&(key))
+#if PYBIND11_INTERNALS_VERSION > 4
+#    define PYBIND11_TLS_KEY_REF Py_tss_t &
+#    if defined(__clang__)
+#        define PYBIND11_TLS_KEY_INIT(var)                                                        \
+            _Pragma("clang diagnostic push")                                         /**/         \
+                _Pragma("clang diagnostic ignored \"-Wmissing-field-initializers\"") /**/         \
+                Py_tss_t var                                                                      \
+                = Py_tss_NEEDS_INIT;                                                              \
+            _Pragma("clang diagnostic pop")
+#    elif defined(__GNUC__) && !defined(__INTEL_COMPILER)
+#        define PYBIND11_TLS_KEY_INIT(var)                                                        \
+            _Pragma("GCC diagnostic push")                                         /**/           \
+                _Pragma("GCC diagnostic ignored \"-Wmissing-field-initializers\"") /**/           \
+                Py_tss_t var                                                                      \
+                = Py_tss_NEEDS_INIT;                                                              \
+            _Pragma("GCC diagnostic pop")
 #    else
-#        define PYBIND11_TLS_KEY_REF Py_tss_t *
-#        define PYBIND11_TLS_KEY_INIT(var) Py_tss_t *var = nullptr;
-#        define PYBIND11_TLS_KEY_CREATE(var)                                                      \
-            (((var) = PyThread_tss_alloc()) != nullptr && (PyThread_tss_create((var)) == 0))
-#        define PYBIND11_TLS_GET_VALUE(key) PyThread_tss_get((key))
-#        define PYBIND11_TLS_REPLACE_VALUE(key, value) PyThread_tss_set((key), (value))
-#        define PYBIND11_TLS_DELETE_VALUE(key) PyThread_tss_set((key), nullptr)
-#        define PYBIND11_TLS_FREE(key) PyThread_tss_free(key)
+#        define PYBIND11_TLS_KEY_INIT(var) Py_tss_t var = Py_tss_NEEDS_INIT;
 #    endif
+#    define PYBIND11_TLS_KEY_CREATE(var) (PyThread_tss_create(&(var)) == 0)
+#    define PYBIND11_TLS_GET_VALUE(key) PyThread_tss_get(&(key))
+#    define PYBIND11_TLS_REPLACE_VALUE(key, value) PyThread_tss_set(&(key), (value))
+#    define PYBIND11_TLS_DELETE_VALUE(key) PyThread_tss_set(&(key), nullptr)
+#    define PYBIND11_TLS_FREE(key) PyThread_tss_delete(&(key))
 #else
-// Usually an int but a long on Cygwin64 with Python 3.x
-#    define PYBIND11_TLS_KEY_REF decltype(PyThread_create_key())
-#    define PYBIND11_TLS_KEY_INIT(var) PYBIND11_TLS_KEY_REF var = 0;
-#    define PYBIND11_TLS_KEY_CREATE(var) (((var) = PyThread_create_key()) != -1)
-#    define PYBIND11_TLS_GET_VALUE(key) PyThread_get_key_value((key))
-#    if defined(PYPY_VERSION)
-// On CPython < 3.4 and on PyPy, `PyThread_set_key_value` strangely does not set
-// the value if it has already been set.  Instead, it must first be deleted and
-// then set again.
-inline void tls_replace_value(PYBIND11_TLS_KEY_REF key, void *value) {
-    PyThread_delete_key_value(key);
-    PyThread_set_key_value(key, value);
-}
-#        define PYBIND11_TLS_DELETE_VALUE(key) PyThread_delete_key_value(key)
-#        define PYBIND11_TLS_REPLACE_VALUE(key, value)                                            \
-            ::pybind11::detail::tls_replace_value((key), (value))
-#    else
-#        define PYBIND11_TLS_DELETE_VALUE(key) PyThread_set_key_value((key), nullptr)
-#        define PYBIND11_TLS_REPLACE_VALUE(key, value) PyThread_set_key_value((key), (value))
-#    endif
-#    define PYBIND11_TLS_FREE(key) (void) key
+#    define PYBIND11_TLS_KEY_REF Py_tss_t *
+#    define PYBIND11_TLS_KEY_INIT(var) Py_tss_t *var = nullptr;
+#    define PYBIND11_TLS_KEY_CREATE(var)                                                          \
+        (((var) = PyThread_tss_alloc()) != nullptr && (PyThread_tss_create((var)) == 0))
+#    define PYBIND11_TLS_GET_VALUE(key) PyThread_tss_get((key))
+#    define PYBIND11_TLS_REPLACE_VALUE(key, value) PyThread_tss_set((key), (value))
+#    define PYBIND11_TLS_DELETE_VALUE(key) PyThread_tss_set((key), nullptr)
+#    define PYBIND11_TLS_FREE(key) PyThread_tss_free(key)
 #endif
 
 // Python loads modules by default with dlopen with the RTLD_LOCAL flag; under libc++ and possibly
@@ -106,7 +107,8 @@ inline void tls_replace_value(PYBIND11_TLS_KEY_REF key, void *value) {
 // libstdc++, this doesn't happen: equality and the type_index hash are based on the type name,
 // which works.  If not under a known-good stl, provide our own name-based hash and equality
 // functions that use the type name.
-#if defined(__GLIBCXX__)
+#if (PYBIND11_INTERNALS_VERSION <= 4 && defined(__GLIBCXX__))                                     \
+    || (PYBIND11_INTERNALS_VERSION >= 5 && !defined(_LIBCPP_VERSION))
 inline bool same_type(const std::type_info &lhs, const std::type_info &rhs) { return lhs == rhs; }
 using type_hash = std::hash<std::type_index>;
 using type_equal_to = std::equal_to<std::type_index>;
@@ -144,15 +146,48 @@ 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 {
+    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(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
+    pymutex mutex;
+#endif
     // std::type_index -> pybind11's type information
     type_map<type_info *> registered_types_cpp;
     // PyTypeObject* -> base type_info(s)
     std::unordered_map<PyTypeObject *, std::vector<type_info *>> registered_types_py;
-    std::unordered_multimap<const void *, instance *> registered_instances; // void * -> instance*
+#ifdef Py_GIL_DISABLED
+    std::unique_ptr<instance_map_shard[]> instance_shards; // void * -> instance*
+    size_t instance_shards_mask;
+#else
+    instance_map registered_instances; // void * -> instance*
+#endif
     std::unordered_set<std::pair<const PyObject *, const char *>, override_hash>
         inactive_override_cache;
     type_map<std::vector<bool (*)(PyObject *, void *&)>> direct_conversions;
@@ -168,16 +203,27 @@ struct internals {
     PyTypeObject *static_property_type;
     PyTypeObject *default_metaclass;
     PyObject *instance_base;
-#if defined(WITH_THREAD)
+    // Unused if PYBIND11_SIMPLE_GIL_MANAGEMENT is defined:
     PYBIND11_TLS_KEY_INIT(tstate)
-#    if PYBIND11_INTERNALS_VERSION > 4
+#if PYBIND11_INTERNALS_VERSION > 4
     PYBIND11_TLS_KEY_INIT(loader_life_support_tls_key)
-#    endif // PYBIND11_INTERNALS_VERSION > 4
+#endif // PYBIND11_INTERNALS_VERSION > 4
+    // Unused if PYBIND11_SIMPLE_GIL_MANAGEMENT is defined:
     PyInterpreterState *istate = nullptr;
+
+#if PYBIND11_INTERNALS_VERSION > 4
+    // Note that we have to use a std::string to allocate memory to ensure a unique address
+    // We want unique addresses since we use pointer equality to compare function records
+    std::string function_record_capsule_name = internals_function_record_capsule_name;
+#endif
+
+    internals() = default;
+    internals(const internals &other) = delete;
+    internals &operator=(const internals &other) = delete;
     ~internals() {
-#    if PYBIND11_INTERNALS_VERSION > 4
+#if PYBIND11_INTERNALS_VERSION > 4
         PYBIND11_TLS_FREE(loader_life_support_tls_key);
-#    endif // PYBIND11_INTERNALS_VERSION > 4
+#endif // PYBIND11_INTERNALS_VERSION > 4
 
         // This destructor is called *after* Py_Finalize() in finalize_interpreter().
         // That *SHOULD BE* fine. The following details what happens when PyThread_tss_free is
@@ -188,7 +234,6 @@ struct internals {
         // that the `tstate` be allocated with the CPython allocator.
         PYBIND11_TLS_FREE(tstate);
     }
-#endif
 };
 
 /// Additional type information which does not fit into the PyTypeObject.
@@ -261,31 +306,30 @@ struct type_info {
 #endif
 
 /// On Linux/OSX, changes in __GXX_ABI_VERSION__ indicate ABI incompatibility.
+/// On MSVC, changes in _MSC_VER may indicate ABI incompatibility (#2898).
 #ifndef PYBIND11_BUILD_ABI
 #    if defined(__GXX_ABI_VERSION)
 #        define PYBIND11_BUILD_ABI "_cxxabi" PYBIND11_TOSTRING(__GXX_ABI_VERSION)
+#    elif defined(_MSC_VER)
+#        define PYBIND11_BUILD_ABI "_mscver" PYBIND11_TOSTRING(_MSC_VER)
 #    else
 #        define PYBIND11_BUILD_ABI ""
 #    endif
 #endif
 
 #ifndef PYBIND11_INTERNALS_KIND
-#    if defined(WITH_THREAD)
-#        define PYBIND11_INTERNALS_KIND ""
-#    else
-#        define PYBIND11_INTERNALS_KIND "_without_thread"
-#    endif
+#    define PYBIND11_INTERNALS_KIND ""
 #endif
 
 #define PYBIND11_INTERNALS_ID                                                                     \
     "__pybind11_internals_v" PYBIND11_TOSTRING(PYBIND11_INTERNALS_VERSION)                        \
-        PYBIND11_INTERNALS_KIND PYBIND11_COMPILER_TYPE PYBIND11_STDLIB PYBIND11_BUILD_ABI         \
-            PYBIND11_BUILD_TYPE "__"
+        PYBIND11_INTERNALS_KIND PYBIND11_COMPILER_TYPE PYBIND11_STDLIB                            \
+            PYBIND11_BUILD_ABI PYBIND11_BUILD_TYPE "__"
 
 #define PYBIND11_MODULE_LOCAL_ID                                                                  \
     "__pybind11_module_local_v" PYBIND11_TOSTRING(PYBIND11_INTERNALS_VERSION)                     \
-        PYBIND11_INTERNALS_KIND PYBIND11_COMPILER_TYPE PYBIND11_STDLIB PYBIND11_BUILD_ABI         \
-            PYBIND11_BUILD_TYPE "__"
+        PYBIND11_INTERNALS_KIND PYBIND11_COMPILER_TYPE PYBIND11_STDLIB                            \
+            PYBIND11_BUILD_ABI PYBIND11_BUILD_TYPE "__"
 
 /// Each module locally stores a pointer to the `internals` data. The data
 /// itself is shared among modules with the same `PYBIND11_INTERNALS_ID`.
@@ -322,7 +366,7 @@ inline bool raise_err(PyObject *exc_type, const char *msg) {
         raise_from(exc_type, msg);
         return true;
     }
-    PyErr_SetString(exc_type, msg);
+    set_error(exc_type, msg);
     return false;
 }
 
@@ -401,6 +445,55 @@ inline void translate_local_exception(std::exception_ptr p) {
 }
 #endif
 
+inline object get_python_state_dict() {
+    object state_dict;
+#if PYBIND11_INTERNALS_VERSION <= 4 || PY_VERSION_HEX < 0x03080000 || defined(PYPY_VERSION)
+    state_dict = reinterpret_borrow<object>(PyEval_GetBuiltins());
+#else
+#    if PY_VERSION_HEX < 0x03090000
+    PyInterpreterState *istate = _PyInterpreterState_Get();
+#    else
+    PyInterpreterState *istate = PyInterpreterState_Get();
+#    endif
+    if (istate) {
+        state_dict = reinterpret_borrow<object>(PyInterpreterState_GetDict(istate));
+    }
+#endif
+    if (!state_dict) {
+        raise_from(PyExc_SystemError, "pybind11::detail::get_python_state_dict() FAILED");
+        throw error_already_set();
+    }
+    return state_dict;
+}
+
+inline object get_internals_obj_from_state_dict(handle state_dict) {
+    return reinterpret_steal<object>(
+        dict_getitemstringref(state_dict.ptr(), PYBIND11_INTERNALS_ID));
+}
+
+inline internals **get_internals_pp_from_capsule(handle obj) {
+    void *raw_ptr = PyCapsule_GetPointer(obj.ptr(), /*name=*/nullptr);
+    if (raw_ptr == nullptr) {
+        raise_from(PyExc_SystemError, "pybind11::detail::get_internals_pp_from_capsule() FAILED");
+        throw error_already_set();
+    }
+    return static_cast<internals **>(raw_ptr);
+}
+
+inline uint64_t round_up_to_next_pow2(uint64_t x) {
+    // Round-up to the next power of two.
+    // See https://graphics.stanford.edu/~seander/bithacks.html#RoundUpPowerOf2
+    x--;
+    x |= (x >> 1);
+    x |= (x >> 2);
+    x |= (x >> 4);
+    x |= (x >> 8);
+    x |= (x >> 16);
+    x |= (x >> 32);
+    x++;
+    return x;
+}
+
 /// Return a reference to the current `internals` data
 PYBIND11_NOINLINE internals &get_internals() {
     auto **&internals_pp = get_internals_pp();
@@ -408,20 +501,27 @@ PYBIND11_NOINLINE internals &get_internals() {
         return **internals_pp;
     }
 
+#if defined(PYBIND11_SIMPLE_GIL_MANAGEMENT)
+    gil_scoped_acquire gil;
+#else
     // Ensure that the GIL is held since we will need to make Python calls.
     // Cannot use py::gil_scoped_acquire here since that constructor calls get_internals.
     struct gil_scoped_acquire_local {
         gil_scoped_acquire_local() : state(PyGILState_Ensure()) {}
+        gil_scoped_acquire_local(const gil_scoped_acquire_local &) = delete;
+        gil_scoped_acquire_local &operator=(const gil_scoped_acquire_local &) = delete;
         ~gil_scoped_acquire_local() { PyGILState_Release(state); }
         const PyGILState_STATE state;
     } gil;
+#endif
+    error_scope err_scope;
 
-    PYBIND11_STR_TYPE id(PYBIND11_INTERNALS_ID);
-    auto builtins = handle(PyEval_GetBuiltins());
-    if (builtins.contains(id) && isinstance<capsule>(builtins[id])) {
-        internals_pp = static_cast<internals **>(capsule(builtins[id]));
-
-        // We loaded builtins through python's builtins, which means that our `error_already_set`
+    dict state_dict = get_python_state_dict();
+    if (object internals_obj = get_internals_obj_from_state_dict(state_dict)) {
+        internals_pp = get_internals_pp_from_capsule(internals_obj);
+    }
+    if (internals_pp && *internals_pp) {
+        // We loaded the internals through `state_dict`, which means that our `error_already_set`
         // and `builtin_exception` may be different local classes than the ones set up in the
         // initial exception translator, below, so add another for our local exception classes.
         //
@@ -437,30 +537,37 @@ PYBIND11_NOINLINE internals &get_internals() {
         }
         auto *&internals_ptr = *internals_pp;
         internals_ptr = new internals();
-#if defined(WITH_THREAD)
 
-#    if PY_VERSION_HEX < 0x03090000
-        PyEval_InitThreads();
-#    endif
         PyThreadState *tstate = PyThreadState_Get();
+        // NOLINTNEXTLINE(bugprone-assignment-in-if-condition)
         if (!PYBIND11_TLS_KEY_CREATE(internals_ptr->tstate)) {
             pybind11_fail("get_internals: could not successfully initialize the tstate TSS key!");
         }
         PYBIND11_TLS_REPLACE_VALUE(internals_ptr->tstate, tstate);
 
-#    if PYBIND11_INTERNALS_VERSION > 4
+#if PYBIND11_INTERNALS_VERSION > 4
+        // NOLINTNEXTLINE(bugprone-assignment-in-if-condition)
         if (!PYBIND11_TLS_KEY_CREATE(internals_ptr->loader_life_support_tls_key)) {
             pybind11_fail("get_internals: could not successfully initialize the "
                           "loader_life_support TSS key!");
         }
-#    endif
-        internals_ptr->istate = tstate->interp;
 #endif
-        builtins[id] = capsule(internals_pp);
+        internals_ptr->istate = tstate->interp;
+        state_dict[PYBIND11_INTERNALS_ID] = capsule(internals_pp);
         internals_ptr->registered_exception_translators.push_front(&translate_exception);
         internals_ptr->static_property_type = make_static_property_type();
         internals_ptr->default_metaclass = make_default_metaclass();
         internals_ptr->instance_base = make_object_base_type(internals_ptr->default_metaclass);
+#ifdef Py_GIL_DISABLED
+        // Scale proportional to the number of cores. 2x is a heuristic to reduce contention.
+        auto num_shards
+            = static_cast<size_t>(round_up_to_next_pow2(2 * std::thread::hardware_concurrency()));
+        if (num_shards == 0) {
+            num_shards = 1;
+        }
+        internals_ptr->instance_shards.reset(new instance_map_shard[num_shards]);
+        internals_ptr->instance_shards_mask = num_shards - 1;
+#endif // Py_GIL_DISABLED
     }
     return **internals_pp;
 }
@@ -474,7 +581,7 @@ PYBIND11_NOINLINE internals &get_internals() {
 struct local_internals {
     type_map<type_info *> registered_types_cpp;
     std::forward_list<ExceptionTranslator> registered_exception_translators;
-#if defined(WITH_THREAD) && PYBIND11_INTERNALS_VERSION == 4
+#if PYBIND11_INTERNALS_VERSION == 4
 
     // For ABI compatibility, we can't store the loader_life_support TLS key in
     // the `internals` struct directly.  Instead, we store it in `shared_data` and
@@ -488,6 +595,7 @@ struct local_internals {
     struct shared_loader_life_support_data {
         PYBIND11_TLS_KEY_INIT(loader_life_support_tls_key)
         shared_loader_life_support_data() {
+            // NOLINTNEXTLINE(bugprone-assignment-in-if-condition)
             if (!PYBIND11_TLS_KEY_CREATE(loader_life_support_tls_key)) {
                 pybind11_fail("local_internals: could not successfully initialize the "
                               "loader_life_support TLS key!");
@@ -506,13 +614,81 @@ struct local_internals {
         loader_life_support_tls_key
             = static_cast<shared_loader_life_support_data *>(ptr)->loader_life_support_tls_key;
     }
-#endif //  defined(WITH_THREAD) && PYBIND11_INTERNALS_VERSION == 4
+#endif //  PYBIND11_INTERNALS_VERSION == 4
 };
 
 /// Works like `get_internals`, but for things which are locally registered.
 inline local_internals &get_local_internals() {
-    static local_internals locals;
-    return locals;
+    // Current static can be created in the interpreter finalization routine. If the later will be
+    // destroyed in another static variable destructor, creation of this static there will cause
+    // static deinitialization fiasco. In order to avoid it we avoid destruction of the
+    // local_internals static. One can read more about the problem and current solution here:
+    // https://google.github.io/styleguide/cppguide.html#Static_and_Global_Variables
+    static auto *locals = new local_internals();
+    return *locals;
+}
+
+#ifdef Py_GIL_DISABLED
+#    define PYBIND11_LOCK_INTERNALS(internals) std::unique_lock<pymutex> lock((internals).mutex)
+#else
+#    define PYBIND11_LOCK_INTERNALS(internals)
+#endif
+
+template <typename F>
+inline auto with_internals(const F &cb) -> decltype(cb(get_internals())) {
+    auto &internals = get_internals();
+    PYBIND11_LOCK_INTERNALS(internals);
+    return cb(internals);
+}
+
+inline std::uint64_t mix64(std::uint64_t z) {
+    // David Stafford's variant 13 of the MurmurHash3 finalizer popularized
+    // by the SplitMix PRNG.
+    // https://zimbry.blogspot.com/2011/09/better-bit-mixing-improving-on.html
+    z = (z ^ (z >> 30)) * 0xbf58476d1ce4e5b9;
+    z = (z ^ (z >> 27)) * 0x94d049bb133111eb;
+    return z ^ (z >> 31);
+}
+
+template <typename F>
+inline auto with_instance_map(const void *ptr,
+                              const F &cb) -> decltype(cb(std::declval<instance_map &>())) {
+    auto &internals = get_internals();
+
+#ifdef Py_GIL_DISABLED
+    // Hash address to compute shard, but ignore low bits. We'd like allocations
+    // from the same thread/core to map to the same shard and allocations from
+    // other threads/cores to map to other shards. Using the high bits is a good
+    // heuristic because memory allocators often have a per-thread
+    // arena/superblock/segment from which smaller allocations are served.
+    auto addr = reinterpret_cast<std::uintptr_t>(ptr);
+    auto hash = mix64(static_cast<std::uint64_t>(addr >> 20));
+    auto idx = static_cast<size_t>(hash & internals.instance_shards_mask);
+
+    auto &shard = internals.instance_shards[idx];
+    std::unique_lock<pymutex> lock(shard.mutex);
+    return cb(shard.registered_instances);
+#else
+    (void) ptr;
+    return cb(internals.registered_instances);
+#endif
+}
+
+// Returns the number of registered instances for testing purposes.  The result may not be
+// consistent if other threads are registering or unregistering instances concurrently.
+inline size_t num_registered_instances() {
+    auto &internals = get_internals();
+#ifdef Py_GIL_DISABLED
+    size_t count = 0;
+    for (size_t i = 0; i <= internals.instance_shards_mask; ++i) {
+        auto &shard = internals.instance_shards[i];
+        std::unique_lock<pymutex> lock(shard.mutex);
+        count += shard.registered_instances.size();
+    }
+    return count;
+#else
+    return internals.registered_instances.size();
+#endif
 }
 
 /// Constructs a std::string with the given arguments, stores it in `internals`, and returns its
@@ -521,26 +697,52 @@ inline local_internals &get_local_internals() {
 /// suitable for c-style strings needed by Python internals (such as PyTypeObject's tp_name).
 template <typename... Args>
 const char *c_str(Args &&...args) {
-    auto &strings = get_internals().static_strings;
+    // GCC 4.8 doesn't like parameter unpack within lambda capture, so use
+    // PYBIND11_LOCK_INTERNALS.
+    auto &internals = get_internals();
+    PYBIND11_LOCK_INTERNALS(internals);
+    auto &strings = internals.static_strings;
     strings.emplace_front(std::forward<Args>(args)...);
     return strings.front().c_str();
 }
 
+inline const char *get_function_record_capsule_name() {
+#if PYBIND11_INTERNALS_VERSION > 4
+    return get_internals().function_record_capsule_name.c_str();
+#else
+    return nullptr;
+#endif
+}
+
+// Determine whether or not the following capsule contains a pybind11 function record.
+// Note that we use `internals` to make sure that only ABI compatible records are touched.
+//
+// This check is currently used in two places:
+// - An important optimization in functional.h to avoid overhead in C++ -> Python -> C++
+// - The sibling feature of cpp_function to allow overloads
+inline bool is_function_record_capsule(const capsule &cap) {
+    // Pointer equality as we rely on internals() to ensure unique pointers
+    return cap.name() == get_function_record_capsule_name();
+}
+
 PYBIND11_NAMESPACE_END(detail)
 
 /// Returns a named pointer that is shared among all extension modules (using the same
 /// pybind11 version) running in the current interpreter. Names starting with underscores
 /// are reserved for internal usage. Returns `nullptr` if no matching entry was found.
 PYBIND11_NOINLINE void *get_shared_data(const std::string &name) {
-    auto &internals = detail::get_internals();
-    auto it = internals.shared_data.find(name);
-    return it != internals.shared_data.end() ? it->second : nullptr;
+    return detail::with_internals([&](detail::internals &internals) {
+        auto it = internals.shared_data.find(name);
+        return it != internals.shared_data.end() ? it->second : nullptr;
+    });
 }
 
 /// Set the shared data that can be later recovered by `get_shared_data()`.
 PYBIND11_NOINLINE void *set_shared_data(const std::string &name, void *data) {
-    detail::get_internals().shared_data[name] = data;
-    return data;
+    return detail::with_internals([&](detail::internals &internals) {
+        internals.shared_data[name] = data;
+        return data;
+    });
 }
 
 /// Returns a typed reference to a shared data entry (by using `get_shared_data()`) if
@@ -548,14 +750,15 @@ PYBIND11_NOINLINE void *set_shared_data(const std::string &name, void *data) {
 /// added to the shared data under the given name and a reference to it is returned.
 template <typename T>
 T &get_or_create_shared_data(const std::string &name) {
-    auto &internals = detail::get_internals();
-    auto it = internals.shared_data.find(name);
-    T *ptr = (T *) (it != internals.shared_data.end() ? it->second : nullptr);
-    if (!ptr) {
-        ptr = new T();
-        internals.shared_data[name] = ptr;
-    }
-    return *ptr;
+    return *detail::with_internals([&](detail::internals &internals) {
+        auto it = internals.shared_data.find(name);
+        T *ptr = (T *) (it != internals.shared_data.end() ? it->second : nullptr);
+        if (!ptr) {
+            ptr = new T();
+            internals.shared_data[name] = ptr;
+        }
+        return ptr;
+    });
 }
 
 PYBIND11_NAMESPACE_END(PYBIND11_NAMESPACE)
diff --git a/include/pybind11/detail/type_caster_base.h b/include/pybind11/detail/type_caster_base.h
index 5b2ec8e1e4..fd8c81b9ac 100644
--- a/include/pybind11/detail/type_caster_base.h
+++ b/include/pybind11/detail/type_caster_base.h
@@ -36,14 +36,13 @@ class loader_life_support {
     loader_life_support *parent = nullptr;
     std::unordered_set<PyObject *> keep_alive;
 
-#if defined(WITH_THREAD)
     // Store stack pointer in thread-local storage.
     static PYBIND11_TLS_KEY_REF get_stack_tls_key() {
-#    if PYBIND11_INTERNALS_VERSION == 4
+#if PYBIND11_INTERNALS_VERSION == 4
         return get_local_internals().loader_life_support_tls_key;
-#    else
+#else
         return get_internals().loader_life_support_tls_key;
-#    endif
+#endif
     }
     static loader_life_support *get_stack_top() {
         return static_cast<loader_life_support *>(PYBIND11_TLS_GET_VALUE(get_stack_tls_key()));
@@ -51,15 +50,6 @@ class loader_life_support {
     static void set_stack_top(loader_life_support *value) {
         PYBIND11_TLS_REPLACE_VALUE(get_stack_tls_key(), value);
     }
-#else
-    // Use single global variable for stack.
-    static loader_life_support **get_stack_pp() {
-        static loader_life_support *global_stack = nullptr;
-        return global_stack;
-    }
-    static loader_life_support *get_stack_top() { return *get_stack_pp(); }
-    static void set_stack_top(loader_life_support *value) { *get_stack_pp() = value; }
-#endif
 
 public:
     /// A new patient frame is created when a function is entered
@@ -102,8 +92,22 @@ class loader_life_support {
 inline std::pair<decltype(internals::registered_types_py)::iterator, bool>
 all_type_info_get_cache(PyTypeObject *type);
 
+// Band-aid workaround to fix a subtle but serious bug in a minimalistic fashion. See PR #4762.
+inline void all_type_info_add_base_most_derived_first(std::vector<type_info *> &bases,
+                                                      type_info *addl_base) {
+    for (auto it = bases.begin(); it != bases.end(); it++) {
+        type_info *existing_base = *it;
+        if (PyType_IsSubtype(addl_base->type, existing_base->type) != 0) {
+            bases.insert(it, addl_base);
+            return;
+        }
+    }
+    bases.push_back(addl_base);
+}
+
 // Populates a just-created cache entry.
 PYBIND11_NOINLINE void all_type_info_populate(PyTypeObject *t, std::vector<type_info *> &bases) {
+    assert(bases.empty());
     std::vector<PyTypeObject *> check;
     for (handle parent : reinterpret_borrow<tuple>(t->tp_bases)) {
         check.push_back((PyTypeObject *) parent.ptr());
@@ -136,7 +140,7 @@ PYBIND11_NOINLINE void all_type_info_populate(PyTypeObject *t, std::vector<type_
                     }
                 }
                 if (!found) {
-                    bases.push_back(tinfo);
+                    all_type_info_add_base_most_derived_first(bases, tinfo);
                 }
             }
         } else if (type->tp_bases) {
@@ -203,12 +207,15 @@ inline detail::type_info *get_local_type_info(const std::type_index &tp) {
 }
 
 inline detail::type_info *get_global_type_info(const std::type_index &tp) {
-    auto &types = get_internals().registered_types_cpp;
-    auto it = types.find(tp);
-    if (it != types.end()) {
-        return it->second;
-    }
-    return nullptr;
+    return with_internals([&](internals &internals) {
+        detail::type_info *type_info = nullptr;
+        auto &types = internals.registered_types_cpp;
+        auto it = types.find(tp);
+        if (it != types.end()) {
+            type_info = it->second;
+        }
+        return type_info;
+    });
 }
 
 /// Return the type info for a given C++ type; on lookup failure can either throw or return
@@ -239,15 +246,17 @@ PYBIND11_NOINLINE handle get_type_handle(const std::type_info &tp, bool throw_if
 // Searches the inheritance graph for a registered Python instance, using all_type_info().
 PYBIND11_NOINLINE handle find_registered_python_instance(void *src,
                                                          const detail::type_info *tinfo) {
-    auto it_instances = get_internals().registered_instances.equal_range(src);
-    for (auto it_i = it_instances.first; it_i != it_instances.second; ++it_i) {
-        for (auto *instance_type : detail::all_type_info(Py_TYPE(it_i->second))) {
-            if (instance_type && same_type(*instance_type->cpptype, *tinfo->cpptype)) {
-                return handle((PyObject *) it_i->second).inc_ref();
+    return with_instance_map(src, [&](instance_map &instances) {
+        auto it_instances = instances.equal_range(src);
+        for (auto it_i = it_instances.first; it_i != it_instances.second; ++it_i) {
+            for (auto *instance_type : detail::all_type_info(Py_TYPE(it_i->second))) {
+                if (instance_type && same_type(*instance_type->cpptype, *tinfo->cpptype)) {
+                    return handle((PyObject *) it_i->second).inc_ref();
+                }
             }
         }
-    }
-    return handle();
+        return handle();
+    });
 }
 
 struct value_and_holder {
@@ -258,9 +267,9 @@ struct value_and_holder {
 
     // 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]} {}
+        : 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;
@@ -322,18 +331,29 @@ struct values_and_holders {
     explicit values_and_holders(instance *inst)
         : inst{inst}, tinfo(all_type_info(Py_TYPE(inst))) {}
 
+    explicit values_and_holders(PyObject *obj)
+        : inst{nullptr}, tinfo(all_type_info(Py_TYPE(obj))) {
+        if (!tinfo.empty()) {
+            inst = reinterpret_cast<instance *>(obj);
+        }
+    }
+
     struct iterator {
     private:
         instance *inst = nullptr;
         const type_vec *types = nullptr;
         value_and_holder curr;
         friend struct values_and_holders;
-        iterator(instance *inst, const type_vec *tinfo)
-            : inst{inst}, types{tinfo},
-              curr(inst /* instance */,
-                   types->empty() ? nullptr : (*types)[0] /* type info */,
-                   0, /* vpos: (non-simple types only): the first vptr comes first */
-                   0 /* index */) {}
+        iterator(instance *inst, const type_vec *tinfo) : inst{inst}, types{tinfo} {
+            if (inst != nullptr) {
+                assert(!types->empty());
+                curr = value_and_holder(
+                    inst /* instance */,
+                    (*types)[0] /* type info */,
+                    0, /* vpos: (non-simple types only): the first vptr comes first */
+                    0 /* index */);
+            }
+        }
         // Past-the-end iterator:
         explicit iterator(size_t end) : curr(end) {}
 
@@ -364,6 +384,16 @@ struct values_and_holders {
     }
 
     size_t size() { return tinfo.size(); }
+
+    // Band-aid workaround to fix a subtle but serious bug in a minimalistic fashion. See PR #4762.
+    bool is_redundant_value_and_holder(const value_and_holder &vh) {
+        for (size_t i = 0; i < vh.index; i++) {
+            if (PyType_IsSubtype(tinfo[i]->type, tinfo[vh.index]->type) != 0) {
+                return true;
+            }
+        }
+        return false;
+    }
 };
 
 /**
@@ -394,15 +424,16 @@ instance::get_value_and_holder(const type_info *find_type /*= nullptr default in
         return value_and_holder();
     }
 
-#if defined(NDEBUG)
-    pybind11_fail("pybind11::detail::instance::get_value_and_holder: "
-                  "type is not a pybind11 base of the given instance "
-                  "(compile in debug mode for type details)");
-#else
+#if defined(PYBIND11_DETAILED_ERROR_MESSAGES)
     pybind11_fail("pybind11::detail::instance::get_value_and_holder: `"
                   + get_fully_qualified_tp_name(find_type->type)
                   + "' is not a pybind11 base of the given `"
                   + get_fully_qualified_tp_name(Py_TYPE(this)) + "' instance");
+#else
+    pybind11_fail(
+        "pybind11::detail::instance::get_value_and_holder: "
+        "type is not a pybind11 base of the given instance "
+        "(#define PYBIND11_DETAILED_ERROR_MESSAGES or compile in debug mode for type details)");
 #endif
 }
 
@@ -469,90 +500,27 @@ PYBIND11_NOINLINE bool isinstance_generic(handle obj, const std::type_info &tp)
     return isinstance(obj, type);
 }
 
-PYBIND11_NOINLINE std::string error_string() {
-    if (!PyErr_Occurred()) {
-        PyErr_SetString(PyExc_RuntimeError, "Unknown internal error occurred");
-        return "Unknown internal error occurred";
-    }
-
-    error_scope scope; // Preserve error state
-
-    std::string errorString;
-    if (scope.type) {
-        errorString += handle(scope.type).attr("__name__").cast<std::string>();
-        errorString += ": ";
-    }
-    if (scope.value) {
-        errorString += (std::string) str(scope.value);
-    }
-
-    PyErr_NormalizeException(&scope.type, &scope.value, &scope.trace);
-
-    if (scope.trace != nullptr) {
-        PyException_SetTraceback(scope.value, scope.trace);
-    }
-
-#if !defined(PYPY_VERSION)
-    if (scope.trace) {
-        auto *trace = (PyTracebackObject *) scope.trace;
-
-        /* Get the deepest trace possible */
-        while (trace->tb_next) {
-            trace = trace->tb_next;
-        }
-
-        PyFrameObject *frame = trace->tb_frame;
-        Py_XINCREF(frame);
-        errorString += "\n\nAt:\n";
-        while (frame) {
-#    if PY_VERSION_HEX >= 0x030900B1
-            PyCodeObject *f_code = PyFrame_GetCode(frame);
-#    else
-            PyCodeObject *f_code = frame->f_code;
-            Py_INCREF(f_code);
-#    endif
-            int lineno = PyFrame_GetLineNumber(frame);
-            errorString += "  ";
-            errorString += handle(f_code->co_filename).cast<std::string>();
-            errorString += '(';
-            errorString += std::to_string(lineno);
-            errorString += "): ";
-            errorString += handle(f_code->co_name).cast<std::string>();
-            errorString += '\n';
-            Py_DECREF(f_code);
-#    if PY_VERSION_HEX >= 0x030900B1
-            auto *b_frame = PyFrame_GetBack(frame);
-#    else
-            auto *b_frame = frame->f_back;
-            Py_XINCREF(b_frame);
-#    endif
-            Py_DECREF(frame);
-            frame = b_frame;
-        }
-    }
-#endif
-
-    return errorString;
-}
-
 PYBIND11_NOINLINE handle get_object_handle(const void *ptr, const detail::type_info *type) {
-    auto &instances = get_internals().registered_instances;
-    auto range = instances.equal_range(ptr);
-    for (auto it = range.first; it != range.second; ++it) {
-        for (const auto &vh : values_and_holders(it->second)) {
-            if (vh.type == type) {
-                return handle((PyObject *) it->second);
+    return with_instance_map(ptr, [&](instance_map &instances) {
+        auto range = instances.equal_range(ptr);
+        for (auto it = range.first; it != range.second; ++it) {
+            for (const auto &vh : values_and_holders(it->second)) {
+                if (vh.type == type) {
+                    return handle((PyObject *) it->second);
+                }
             }
         }
-    }
-    return handle();
+        return handle();
+    });
 }
 
 inline PyThreadState *get_thread_state_unchecked() {
 #if defined(PYPY_VERSION)
     return PyThreadState_GET();
-#else
+#elif PY_VERSION_HEX < 0x030D0000
     return _PyThreadState_UncheckedGet();
+#else
+    return PyThreadState_GetUnchecked();
 #endif
 }
 
@@ -612,14 +580,15 @@ class type_caster_generic {
                 if (copy_constructor) {
                     valueptr = copy_constructor(src);
                 } else {
-#if defined(NDEBUG)
-                    throw cast_error("return_value_policy = copy, but type is "
-                                     "non-copyable! (compile in debug mode for details)");
-#else
+#if defined(PYBIND11_DETAILED_ERROR_MESSAGES)
                     std::string type_name(tinfo->cpptype->name());
                     detail::clean_type_id(type_name);
                     throw cast_error("return_value_policy = copy, but type " + type_name
                                      + " is non-copyable!");
+#else
+                    throw cast_error("return_value_policy = copy, but type is "
+                                     "non-copyable! (#define PYBIND11_DETAILED_ERROR_MESSAGES or "
+                                     "compile in debug mode for details)");
 #endif
                 }
                 wrapper->owned = true;
@@ -631,15 +600,16 @@ class type_caster_generic {
                 } else if (copy_constructor) {
                     valueptr = copy_constructor(src);
                 } else {
-#if defined(NDEBUG)
-                    throw cast_error("return_value_policy = move, but type is neither "
-                                     "movable nor copyable! "
-                                     "(compile in debug mode for details)");
-#else
+#if defined(PYBIND11_DETAILED_ERROR_MESSAGES)
                     std::string type_name(tinfo->cpptype->name());
                     detail::clean_type_id(type_name);
                     throw cast_error("return_value_policy = move, but type " + type_name
                                      + " is neither movable nor copyable!");
+#else
+                    throw cast_error("return_value_policy = move, but type is neither "
+                                     "movable nor copyable! "
+                                     "(#define PYBIND11_DETAILED_ERROR_MESSAGES or compile in "
+                                     "debug mode for details)");
 #endif
                 }
                 wrapper->owned = true;
@@ -849,7 +819,7 @@ class type_caster_generic {
         std::string tname = rtti_type ? rtti_type->name() : cast_type.name();
         detail::clean_type_id(tname);
         std::string msg = "Unregistered type : " + tname;
-        PyErr_SetString(PyExc_TypeError, msg.c_str());
+        set_error(PyExc_TypeError, msg.c_str());
         return {nullptr, nullptr};
     }
 
@@ -885,23 +855,179 @@ using movable_cast_op_type
                                   typename std::add_rvalue_reference<intrinsic_t<T>>::type,
                                   typename std::add_lvalue_reference<intrinsic_t<T>>::type>>;
 
-// std::is_copy_constructible isn't quite enough: it lets std::vector<T> (and similar) through when
-// T is non-copyable, but code containing such a copy constructor fails to actually compile.
-template <typename T, typename SFINAE = void>
-struct is_copy_constructible : std::is_copy_constructible<T> {};
+// Does the container have a mapped type and is it recursive?
+// Implemented by specializations below.
+template <typename Container, typename SFINAE = void>
+struct container_mapped_type_traits {
+    static constexpr bool has_mapped_type = false;
+    static constexpr bool has_recursive_mapped_type = false;
+};
 
-// Specialization for types that appear to be copy constructible but also look like stl containers
-// (we specifically check for: has `value_type` and `reference` with `reference = value_type&`): if
-// so, copy constructability depends on whether the value_type is copy constructible.
 template <typename Container>
-struct is_copy_constructible<
+struct container_mapped_type_traits<
     Container,
-    enable_if_t<
-        all_of<std::is_copy_constructible<Container>,
-               std::is_same<typename Container::value_type &, typename Container::reference>,
-               // Avoid infinite recursion
-               negation<std::is_same<Container, typename Container::value_type>>>::value>>
-    : is_copy_constructible<typename Container::value_type> {};
+    typename std::enable_if<
+        std::is_same<typename Container::mapped_type, Container>::value>::type> {
+    static constexpr bool has_mapped_type = true;
+    static constexpr bool has_recursive_mapped_type = true;
+};
+
+template <typename Container>
+struct container_mapped_type_traits<
+    Container,
+    typename std::enable_if<
+        negation<std::is_same<typename Container::mapped_type, Container>>::value>::type> {
+    static constexpr bool has_mapped_type = true;
+    static constexpr bool has_recursive_mapped_type = false;
+};
+
+// Does the container have a value type and is it recursive?
+// Implemented by specializations below.
+template <typename Container, typename SFINAE = void>
+struct container_value_type_traits : std::false_type {
+    static constexpr bool has_value_type = false;
+    static constexpr bool has_recursive_value_type = false;
+};
+
+template <typename Container>
+struct container_value_type_traits<
+    Container,
+    typename std::enable_if<
+        std::is_same<typename Container::value_type, Container>::value>::type> {
+    static constexpr bool has_value_type = true;
+    static constexpr bool has_recursive_value_type = true;
+};
+
+template <typename Container>
+struct container_value_type_traits<
+    Container,
+    typename std::enable_if<
+        negation<std::is_same<typename Container::value_type, Container>>::value>::type> {
+    static constexpr bool has_value_type = true;
+    static constexpr bool has_recursive_value_type = false;
+};
+
+/*
+ * Tag to be used for representing the bottom of recursively defined types.
+ * Define this tag so we don't have to use void.
+ */
+struct recursive_bottom {};
+
+/*
+ * Implementation detail of `recursive_container_traits` below.
+ * `T` is the `value_type` of the container, which might need to be modified to
+ * avoid recursive types and const types.
+ */
+template <typename T, bool is_this_a_map>
+struct impl_type_to_check_recursively {
+    /*
+     * If the container is recursive, then no further recursion should be done.
+     */
+    using if_recursive = recursive_bottom;
+    /*
+     * Otherwise yield `T` unchanged.
+     */
+    using if_not_recursive = T;
+};
+
+/*
+ * For pairs - only as value type of a map -, the first type should remove the `const`.
+ * Also, if the map is recursive, then the recursive checking should consider
+ * the first type only.
+ */
+template <typename A, typename B>
+struct impl_type_to_check_recursively<std::pair<A, B>, /* is_this_a_map = */ true> {
+    using if_recursive = typename std::remove_const<A>::type;
+    using if_not_recursive = std::pair<typename std::remove_const<A>::type, B>;
+};
+
+/*
+ * Implementation of `recursive_container_traits` below.
+ */
+template <typename Container, typename SFINAE = void>
+struct impl_recursive_container_traits {
+    using type_to_check_recursively = recursive_bottom;
+};
+
+template <typename Container>
+struct impl_recursive_container_traits<
+    Container,
+    typename std::enable_if<container_value_type_traits<Container>::has_value_type>::type> {
+    static constexpr bool is_recursive
+        = container_mapped_type_traits<Container>::has_recursive_mapped_type
+          || container_value_type_traits<Container>::has_recursive_value_type;
+    /*
+     * This member dictates which type Pybind11 should check recursively in traits
+     * such as `is_move_constructible`, `is_copy_constructible`, `is_move_assignable`, ...
+     * Direct access to `value_type` should be avoided:
+     * 1. `value_type` might recursively contain the type again
+     * 2. `value_type` of STL map types is `std::pair<A const, B>`, the `const`
+     *    should be removed.
+     *
+     */
+    using type_to_check_recursively = typename std::conditional<
+        is_recursive,
+        typename impl_type_to_check_recursively<
+            typename Container::value_type,
+            container_mapped_type_traits<Container>::has_mapped_type>::if_recursive,
+        typename impl_type_to_check_recursively<
+            typename Container::value_type,
+            container_mapped_type_traits<Container>::has_mapped_type>::if_not_recursive>::type;
+};
+
+/*
+ * This trait defines the `type_to_check_recursively` which is needed to properly
+ * handle recursively defined traits such as `is_move_constructible` without going
+ * into an infinite recursion.
+ * Should be used instead of directly accessing the `value_type`.
+ * It cancels the recursion by returning the `recursive_bottom` tag.
+ *
+ * The default definition of `type_to_check_recursively` is as follows:
+ *
+ * 1. By default, it is `recursive_bottom`, so that the recursion is canceled.
+ * 2. If the type is non-recursive and defines a `value_type`, then the `value_type` is used.
+ *    If the `value_type` is a pair and a `mapped_type` is defined,
+ *    then the `const` is removed from the first type.
+ * 3. If the type is recursive and `value_type` is not a pair, then `recursive_bottom` is returned.
+ * 4. If the type is recursive and `value_type` is a pair and a `mapped_type` is defined,
+ *    then `const` is removed from the first type and the first type is returned.
+ *
+ * This behavior can be extended by the user as seen in test_stl_binders.cpp.
+ *
+ * This struct is exactly the same as impl_recursive_container_traits.
+ * The duplication achieves that user-defined specializations don't compete
+ * with internal specializations, but take precedence.
+ */
+template <typename Container, typename SFINAE = void>
+struct recursive_container_traits : impl_recursive_container_traits<Container> {};
+
+template <typename T>
+struct is_move_constructible
+    : all_of<std::is_move_constructible<T>,
+             is_move_constructible<
+                 typename recursive_container_traits<T>::type_to_check_recursively>> {};
+
+template <>
+struct is_move_constructible<recursive_bottom> : std::true_type {};
+
+// Likewise for std::pair
+// (after C++17 it is mandatory that the move constructor not exist when the two types aren't
+// themselves move constructible, but this can not be relied upon when T1 or T2 are themselves
+// containers).
+template <typename T1, typename T2>
+struct is_move_constructible<std::pair<T1, T2>>
+    : all_of<is_move_constructible<T1>, is_move_constructible<T2>> {};
+
+// std::is_copy_constructible isn't quite enough: it lets std::vector<T> (and similar) through when
+// T is non-copyable, but code containing such a copy constructor fails to actually compile.
+template <typename T>
+struct is_copy_constructible
+    : all_of<std::is_copy_constructible<T>,
+             is_copy_constructible<
+                 typename recursive_container_traits<T>::type_to_check_recursively>> {};
+
+template <>
+struct is_copy_constructible<recursive_bottom> : std::true_type {};
 
 // Likewise for std::pair
 // (after C++17 it is mandatory that the copy constructor not exist when the two types aren't
@@ -912,14 +1038,16 @@ struct is_copy_constructible<std::pair<T1, T2>>
     : all_of<is_copy_constructible<T1>, is_copy_constructible<T2>> {};
 
 // The same problems arise with std::is_copy_assignable, so we use the same workaround.
-template <typename T, typename SFINAE = void>
-struct is_copy_assignable : std::is_copy_assignable<T> {};
-template <typename Container>
-struct is_copy_assignable<Container,
-                          enable_if_t<all_of<std::is_copy_assignable<Container>,
-                                             std::is_same<typename Container::value_type &,
-                                                          typename Container::reference>>::value>>
-    : is_copy_assignable<typename Container::value_type> {};
+template <typename T>
+struct is_copy_assignable
+    : all_of<
+          std::is_copy_assignable<T>,
+          is_copy_assignable<typename recursive_container_traits<T>::type_to_check_recursively>> {
+};
+
+template <>
+struct is_copy_assignable<recursive_bottom> : std::true_type {};
+
 template <typename T1, typename T2>
 struct is_copy_assignable<std::pair<T1, T2>>
     : all_of<is_copy_assignable<T1>, is_copy_assignable<T2>> {};
@@ -979,11 +1107,11 @@ class type_caster_base : public type_caster_generic {
             || policy == return_value_policy::automatic_reference) {
             policy = return_value_policy::copy;
         }
-        return cast(&src, policy, parent);
+        return cast(std::addressof(src), policy, parent);
     }
 
     static handle cast(itype &&src, return_value_policy, handle parent) {
-        return cast(&src, return_value_policy::move, parent);
+        return cast(std::addressof(src), return_value_policy::move, parent);
     }
 
     // Returns a (pointer, type_info) pair taking care of necessary type lookup for a
@@ -1052,14 +1180,14 @@ class type_caster_base : public type_caster_generic {
        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>()), Constructor{}) {
+    static auto make_copy_constructor(const T *) -> 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<std::is_move_constructible<T>::value>>
-    static auto make_move_constructor(const T *)
-        -> decltype(new T(std::declval<T &&>()), Constructor{}) {
+    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 &&>()),
+                                                             Constructor{}) {
         return [](const void *arg) -> void * {
             return new T(std::move(*const_cast<T *>(reinterpret_cast<const T *>(arg))));
         };
@@ -1069,5 +1197,18 @@ class type_caster_base : public type_caster_generic {
     static Constructor make_move_constructor(...) { return nullptr; }
 };
 
+inline std::string quote_cpp_type_name(const std::string &cpp_type_name) {
+    return cpp_type_name; // No-op for now. See PR #4888
+}
+
+PYBIND11_NOINLINE std::string type_info_description(const std::type_info &ti) {
+    if (auto *type_data = get_type_info(ti)) {
+        handle th((PyObject *) type_data->type);
+        return th.attr("__module__").cast<std::string>() + '.'
+               + th.attr("__qualname__").cast<std::string>();
+    }
+    return quote_cpp_type_name(clean_type_id(ti.name()));
+}
+
 PYBIND11_NAMESPACE_END(detail)
 PYBIND11_NAMESPACE_END(PYBIND11_NAMESPACE)
diff --git a/include/pybind11/detail/typeid.h b/include/pybind11/detail/typeid.h
index 8d99fc0286..a67b52135b 100644
--- a/include/pybind11/detail/typeid.h
+++ b/include/pybind11/detail/typeid.h
@@ -20,6 +20,7 @@
 
 PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)
 PYBIND11_NAMESPACE_BEGIN(detail)
+
 /// Erase all occurrences of a substring
 inline void erase_all(std::string &string, const std::string &search) {
     for (size_t pos = 0;;) {
@@ -46,14 +47,19 @@ PYBIND11_NOINLINE void clean_type_id(std::string &name) {
 #endif
     detail::erase_all(name, "pybind11::");
 }
+
+inline std::string clean_type_id(const char *typeid_name) {
+    std::string name(typeid_name);
+    detail::clean_type_id(name);
+    return name;
+}
+
 PYBIND11_NAMESPACE_END(detail)
 
 /// Return a string representation of a C++ type
 template <typename T>
 static std::string type_id() {
-    std::string name(typeid(T).name());
-    detail::clean_type_id(name);
-    return name;
+    return detail::clean_type_id(typeid(T).name());
 }
 
 PYBIND11_NAMESPACE_END(PYBIND11_NAMESPACE)
diff --git a/include/pybind11/eigen.h b/include/pybind11/eigen.h
index f658168ded..273b9c9308 100644
--- a/include/pybind11/eigen.h
+++ b/include/pybind11/eigen.h
@@ -9,694 +9,4 @@
 
 #pragma once
 
-/* HINT: To suppress warnings originating from the Eigen headers, use -isystem.
-   See also:
-       https://stackoverflow.com/questions/2579576/i-dir-vs-isystem-dir
-       https://stackoverflow.com/questions/1741816/isystem-for-ms-visual-studio-c-compiler
-*/
-
-#include "numpy.h"
-
-// The C4127 suppression was introduced for Eigen 3.4.0. In theory we could
-// make it version specific, or even remove it later, but considering that
-// 1. C4127 is generally far more distracting than useful for modern template code, and
-// 2. we definitely want to ignore any MSVC warnings originating from Eigen code,
-//    it is probably best to keep this around indefinitely.
-#if defined(_MSC_VER)
-#    pragma warning(push)
-#    pragma warning(disable : 4127) // C4127: conditional expression is constant
-#    pragma warning(disable : 5054) // https://github.com/pybind/pybind11/pull/3741
-//       C5054: operator '&': deprecated between enumerations of different types
-#endif
-
-#include <Eigen/Core>
-#include <Eigen/SparseCore>
-
-#if defined(_MSC_VER)
-#    pragma warning(pop)
-#endif
-
-// Eigen prior to 3.2.7 doesn't have proper move constructors--but worse, some classes get implicit
-// move constructors that break things.  We could detect this an explicitly copy, but an extra copy
-// of matrices seems highly undesirable.
-static_assert(EIGEN_VERSION_AT_LEAST(3, 2, 7),
-              "Eigen support in pybind11 requires Eigen >= 3.2.7");
-
-PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)
-
-// Provide a convenience alias for easier pass-by-ref usage with fully dynamic strides:
-using EigenDStride = Eigen::Stride<Eigen::Dynamic, Eigen::Dynamic>;
-template <typename MatrixType>
-using EigenDRef = Eigen::Ref<MatrixType, 0, EigenDStride>;
-template <typename MatrixType>
-using EigenDMap = Eigen::Map<MatrixType, 0, EigenDStride>;
-
-PYBIND11_NAMESPACE_BEGIN(detail)
-
-#if EIGEN_VERSION_AT_LEAST(3, 3, 0)
-using EigenIndex = Eigen::Index;
-template <typename Scalar, int Flags, typename StorageIndex>
-using EigenMapSparseMatrix = Eigen::Map<Eigen::SparseMatrix<Scalar, Flags, StorageIndex>>;
-#else
-using EigenIndex = EIGEN_DEFAULT_DENSE_INDEX_TYPE;
-template <typename Scalar, int Flags, typename StorageIndex>
-using EigenMapSparseMatrix = Eigen::MappedSparseMatrix<Scalar, Flags, StorageIndex>;
-#endif
-
-// Matches Eigen::Map, Eigen::Ref, blocks, etc:
-template <typename T>
-using is_eigen_dense_map = all_of<is_template_base_of<Eigen::DenseBase, T>,
-                                  std::is_base_of<Eigen::MapBase<T, Eigen::ReadOnlyAccessors>, T>>;
-template <typename T>
-using is_eigen_mutable_map = std::is_base_of<Eigen::MapBase<T, Eigen::WriteAccessors>, T>;
-template <typename T>
-using is_eigen_dense_plain
-    = all_of<negation<is_eigen_dense_map<T>>, is_template_base_of<Eigen::PlainObjectBase, T>>;
-template <typename T>
-using is_eigen_sparse = is_template_base_of<Eigen::SparseMatrixBase, T>;
-// Test for objects inheriting from EigenBase<Derived> that aren't captured by the above.  This
-// basically covers anything that can be assigned to a dense matrix but that don't have a typical
-// matrix data layout that can be copied from their .data().  For example, DiagonalMatrix and
-// SelfAdjointView fall into this category.
-template <typename T>
-using is_eigen_other
-    = all_of<is_template_base_of<Eigen::EigenBase, T>,
-             negation<any_of<is_eigen_dense_map<T>, is_eigen_dense_plain<T>, is_eigen_sparse<T>>>>;
-
-// Captures numpy/eigen conformability status (returned by EigenProps::conformable()):
-template <bool EigenRowMajor>
-struct EigenConformable {
-    bool conformable = false;
-    EigenIndex rows = 0, cols = 0;
-    EigenDStride stride{0, 0};    // Only valid if negativestrides is false!
-    bool negativestrides = false; // If true, do not use stride!
-
-    // NOLINTNEXTLINE(google-explicit-constructor)
-    EigenConformable(bool fits = false) : conformable{fits} {}
-    // Matrix type:
-    EigenConformable(EigenIndex r, EigenIndex c, EigenIndex rstride, EigenIndex cstride)
-        : conformable{true}, rows{r}, cols{c},
-          // TODO: when Eigen bug #747 is fixed, remove the tests for non-negativity.
-          // http://eigen.tuxfamily.org/bz/show_bug.cgi?id=747
-          stride{EigenRowMajor ? (rstride > 0 ? rstride : 0)
-                               : (cstride > 0 ? cstride : 0) /* outer stride */,
-                 EigenRowMajor ? (cstride > 0 ? cstride : 0)
-                               : (rstride > 0 ? rstride : 0) /* inner stride */},
-          negativestrides{rstride < 0 || cstride < 0} {}
-    // Vector type:
-    EigenConformable(EigenIndex r, EigenIndex c, EigenIndex stride)
-        : EigenConformable(r, c, r == 1 ? c * stride : stride, c == 1 ? r : r * stride) {}
-
-    template <typename props>
-    bool stride_compatible() const {
-        // To have compatible strides, we need (on both dimensions) one of fully dynamic strides,
-        // matching strides, or a dimension size of 1 (in which case the stride value is
-        // irrelevant)
-        return !negativestrides
-               && (props::inner_stride == Eigen::Dynamic || props::inner_stride == stride.inner()
-                   || (EigenRowMajor ? cols : rows) == 1)
-               && (props::outer_stride == Eigen::Dynamic || props::outer_stride == stride.outer()
-                   || (EigenRowMajor ? rows : cols) == 1);
-    }
-    // NOLINTNEXTLINE(google-explicit-constructor)
-    operator bool() const { return conformable; }
-};
-
-template <typename Type>
-struct eigen_extract_stride {
-    using type = Type;
-};
-template <typename PlainObjectType, int MapOptions, typename StrideType>
-struct eigen_extract_stride<Eigen::Map<PlainObjectType, MapOptions, StrideType>> {
-    using type = StrideType;
-};
-template <typename PlainObjectType, int Options, typename StrideType>
-struct eigen_extract_stride<Eigen::Ref<PlainObjectType, Options, StrideType>> {
-    using type = StrideType;
-};
-
-// Helper struct for extracting information from an Eigen type
-template <typename Type_>
-struct EigenProps {
-    using Type = Type_;
-    using Scalar = typename Type::Scalar;
-    using StrideType = typename eigen_extract_stride<Type>::type;
-    static constexpr EigenIndex rows = Type::RowsAtCompileTime, cols = Type::ColsAtCompileTime,
-                                size = Type::SizeAtCompileTime;
-    static constexpr bool row_major = Type::IsRowMajor,
-                          vector
-                          = Type::IsVectorAtCompileTime, // At least one dimension has fixed size 1
-        fixed_rows = rows != Eigen::Dynamic, fixed_cols = cols != Eigen::Dynamic,
-                          fixed = size != Eigen::Dynamic, // Fully-fixed size
-        dynamic = !fixed_rows && !fixed_cols;             // Fully-dynamic size
-
-    template <EigenIndex i, EigenIndex ifzero>
-    using if_zero = std::integral_constant<EigenIndex, i == 0 ? ifzero : i>;
-    static constexpr EigenIndex inner_stride
-        = if_zero<StrideType::InnerStrideAtCompileTime, 1>::value,
-        outer_stride = if_zero < StrideType::OuterStrideAtCompileTime,
-        vector      ? size
-        : row_major ? cols
-                    : rows > ::value;
-    static constexpr bool dynamic_stride
-        = inner_stride == Eigen::Dynamic && outer_stride == Eigen::Dynamic;
-    static constexpr bool requires_row_major
-        = !dynamic_stride && !vector && (row_major ? inner_stride : outer_stride) == 1;
-    static constexpr bool requires_col_major
-        = !dynamic_stride && !vector && (row_major ? outer_stride : inner_stride) == 1;
-
-    // Takes an input array and determines whether we can make it fit into the Eigen type.  If
-    // the array is a vector, we attempt to fit it into either an Eigen 1xN or Nx1 vector
-    // (preferring the latter if it will fit in either, i.e. for a fully dynamic matrix type).
-    static EigenConformable<row_major> conformable(const array &a) {
-        const auto dims = a.ndim();
-        if (dims < 1 || dims > 2) {
-            return false;
-        }
-
-        if (dims == 2) { // Matrix type: require exact match (or dynamic)
-
-            EigenIndex np_rows = a.shape(0), np_cols = a.shape(1),
-                       np_rstride = a.strides(0) / static_cast<ssize_t>(sizeof(Scalar)),
-                       np_cstride = a.strides(1) / static_cast<ssize_t>(sizeof(Scalar));
-            if ((PYBIND11_SILENCE_MSVC_C4127(fixed_rows) && np_rows != rows)
-                || (PYBIND11_SILENCE_MSVC_C4127(fixed_cols) && np_cols != cols)) {
-                return false;
-            }
-
-            return {np_rows, np_cols, np_rstride, np_cstride};
-        }
-
-        // Otherwise we're storing an n-vector.  Only one of the strides will be used, but
-        // whichever is used, we want the (single) numpy stride value.
-        const EigenIndex n = a.shape(0),
-                         stride = a.strides(0) / static_cast<ssize_t>(sizeof(Scalar));
-
-        if (vector) { // Eigen type is a compile-time vector
-            if (PYBIND11_SILENCE_MSVC_C4127(fixed) && size != n) {
-                return false; // Vector size mismatch
-            }
-            return {rows == 1 ? 1 : n, cols == 1 ? 1 : n, stride};
-        }
-        if (fixed) {
-            // The type has a fixed size, but is not a vector: abort
-            return false;
-        }
-        if (fixed_cols) {
-            // Since this isn't a vector, cols must be != 1.  We allow this only if it exactly
-            // equals the number of elements (rows is Dynamic, and so 1 row is allowed).
-            if (cols != n) {
-                return false;
-            }
-            return {1, n, stride};
-        } // Otherwise it's either fully dynamic, or column dynamic; both become a column vector
-        if (PYBIND11_SILENCE_MSVC_C4127(fixed_rows) && rows != n) {
-            return false;
-        }
-        return {n, 1, stride};
-    }
-
-    static constexpr bool show_writeable
-        = is_eigen_dense_map<Type>::value && is_eigen_mutable_map<Type>::value;
-    static constexpr bool show_order = is_eigen_dense_map<Type>::value;
-    static constexpr bool show_c_contiguous = show_order && requires_row_major;
-    static constexpr bool show_f_contiguous
-        = !show_c_contiguous && show_order && requires_col_major;
-
-    static constexpr auto descriptor
-        = const_name("numpy.ndarray[") + npy_format_descriptor<Scalar>::name + const_name("[")
-          + const_name<fixed_rows>(const_name<(size_t) rows>(), const_name("m")) + const_name(", ")
-          + const_name<fixed_cols>(const_name<(size_t) cols>(), const_name("n")) + const_name("]")
-          +
-          // For a reference type (e.g. Ref<MatrixXd>) we have other constraints that might need to
-          // be satisfied: writeable=True (for a mutable reference), and, depending on the map's
-          // stride options, possibly f_contiguous or c_contiguous.  We include them in the
-          // descriptor output to provide some hint as to why a TypeError is occurring (otherwise
-          // it can be confusing to see that a function accepts a 'numpy.ndarray[float64[3,2]]' and
-          // an error message that you *gave* a numpy.ndarray of the right type and dimensions.
-          const_name<show_writeable>(", flags.writeable", "")
-          + const_name<show_c_contiguous>(", flags.c_contiguous", "")
-          + const_name<show_f_contiguous>(", flags.f_contiguous", "") + const_name("]");
-};
-
-// Casts an Eigen type to numpy array.  If given a base, the numpy array references the src data,
-// otherwise it'll make a copy.  writeable lets you turn off the writeable flag for the array.
-template <typename props>
-handle
-eigen_array_cast(typename props::Type const &src, handle base = handle(), bool writeable = true) {
-    constexpr ssize_t elem_size = sizeof(typename props::Scalar);
-    array a;
-    if (props::vector) {
-        a = array({src.size()}, {elem_size * src.innerStride()}, src.data(), base);
-    } else {
-        a = array({src.rows(), src.cols()},
-                  {elem_size * src.rowStride(), elem_size * src.colStride()},
-                  src.data(),
-                  base);
-    }
-
-    if (!writeable) {
-        array_proxy(a.ptr())->flags &= ~detail::npy_api::NPY_ARRAY_WRITEABLE_;
-    }
-
-    return a.release();
-}
-
-// Takes an lvalue ref to some Eigen type and a (python) base object, creating a numpy array that
-// reference the Eigen object's data with `base` as the python-registered base class (if omitted,
-// the base will be set to None, and lifetime management is up to the caller).  The numpy array is
-// non-writeable if the given type is const.
-template <typename props, typename Type>
-handle eigen_ref_array(Type &src, handle parent = none()) {
-    // none here is to get past array's should-we-copy detection, which currently always
-    // copies when there is no base.  Setting the base to None should be harmless.
-    return eigen_array_cast<props>(src, parent, !std::is_const<Type>::value);
-}
-
-// Takes a pointer to some dense, plain Eigen type, builds a capsule around it, then returns a
-// numpy array that references the encapsulated data with a python-side reference to the capsule to
-// tie its destruction to that of any dependent python objects.  Const-ness is determined by
-// whether or not the Type of the pointer given is const.
-template <typename props, typename Type, typename = enable_if_t<is_eigen_dense_plain<Type>::value>>
-handle eigen_encapsulate(Type *src) {
-    capsule base(src, [](void *o) { delete static_cast<Type *>(o); });
-    return eigen_ref_array<props>(*src, base);
-}
-
-// Type caster for regular, dense matrix types (e.g. MatrixXd), but not maps/refs/etc. of dense
-// types.
-template <typename Type>
-struct type_caster<Type, enable_if_t<is_eigen_dense_plain<Type>::value>> {
-    using Scalar = typename Type::Scalar;
-    using props = EigenProps<Type>;
-
-    bool load(handle src, bool convert) {
-        // If we're in no-convert mode, only load if given an array of the correct type
-        if (!convert && !isinstance<array_t<Scalar>>(src)) {
-            return false;
-        }
-
-        // Coerce into an array, but don't do type conversion yet; the copy below handles it.
-        auto buf = array::ensure(src);
-
-        if (!buf) {
-            return false;
-        }
-
-        auto dims = buf.ndim();
-        if (dims < 1 || dims > 2) {
-            return false;
-        }
-
-        auto fits = props::conformable(buf);
-        if (!fits) {
-            return false;
-        }
-
-        // Allocate the new type, then build a numpy reference into it
-        value = Type(fits.rows, fits.cols);
-        auto ref = reinterpret_steal<array>(eigen_ref_array<props>(value));
-        if (dims == 1) {
-            ref = ref.squeeze();
-        } else if (ref.ndim() == 1) {
-            buf = buf.squeeze();
-        }
-
-        int result = detail::npy_api::get().PyArray_CopyInto_(ref.ptr(), buf.ptr());
-
-        if (result < 0) { // Copy failed!
-            PyErr_Clear();
-            return false;
-        }
-
-        return true;
-    }
-
-private:
-    // Cast implementation
-    template <typename CType>
-    static handle cast_impl(CType *src, return_value_policy policy, handle parent) {
-        switch (policy) {
-            case return_value_policy::take_ownership:
-            case return_value_policy::automatic:
-                return eigen_encapsulate<props>(src);
-            case return_value_policy::move:
-                return eigen_encapsulate<props>(new CType(std::move(*src)));
-            case return_value_policy::copy:
-                return eigen_array_cast<props>(*src);
-            case return_value_policy::reference:
-            case return_value_policy::automatic_reference:
-                return eigen_ref_array<props>(*src);
-            case return_value_policy::reference_internal:
-                return eigen_ref_array<props>(*src, parent);
-            default:
-                throw cast_error("unhandled return_value_policy: should not happen!");
-        };
-    }
-
-public:
-    // Normal returned non-reference, non-const value:
-    static handle cast(Type &&src, return_value_policy /* policy */, handle parent) {
-        return cast_impl(&src, return_value_policy::move, parent);
-    }
-    // If you return a non-reference const, we mark the numpy array readonly:
-    static handle cast(const Type &&src, return_value_policy /* policy */, handle parent) {
-        return cast_impl(&src, return_value_policy::move, parent);
-    }
-    // lvalue reference return; default (automatic) becomes copy
-    static handle cast(Type &src, return_value_policy policy, handle parent) {
-        if (policy == return_value_policy::automatic
-            || policy == return_value_policy::automatic_reference) {
-            policy = return_value_policy::copy;
-        }
-        return cast_impl(&src, policy, parent);
-    }
-    // const lvalue reference return; default (automatic) becomes copy
-    static handle cast(const Type &src, return_value_policy policy, handle parent) {
-        if (policy == return_value_policy::automatic
-            || policy == return_value_policy::automatic_reference) {
-            policy = return_value_policy::copy;
-        }
-        return cast(&src, policy, parent);
-    }
-    // non-const pointer return
-    static handle cast(Type *src, return_value_policy policy, handle parent) {
-        return cast_impl(src, policy, parent);
-    }
-    // const pointer return
-    static handle cast(const Type *src, return_value_policy policy, handle parent) {
-        return cast_impl(src, policy, parent);
-    }
-
-    static constexpr auto name = props::descriptor;
-
-    // NOLINTNEXTLINE(google-explicit-constructor)
-    operator Type *() { return &value; }
-    // NOLINTNEXTLINE(google-explicit-constructor)
-    operator Type &() { return value; }
-    // NOLINTNEXTLINE(google-explicit-constructor)
-    operator Type &&() && { return std::move(value); }
-    template <typename T>
-    using cast_op_type = movable_cast_op_type<T>;
-
-private:
-    Type value;
-};
-
-// Base class for casting reference/map/block/etc. objects back to python.
-template <typename MapType>
-struct eigen_map_caster {
-private:
-    using props = EigenProps<MapType>;
-
-public:
-    // Directly referencing a ref/map's data is a bit dangerous (whatever the map/ref points to has
-    // to stay around), but we'll allow it under the assumption that you know what you're doing
-    // (and have an appropriate keep_alive in place).  We return a numpy array pointing directly at
-    // the ref's data (The numpy array ends up read-only if the ref was to a const matrix type.)
-    // Note that this means you need to ensure you don't destroy the object in some other way (e.g.
-    // with an appropriate keep_alive, or with a reference to a statically allocated matrix).
-    static handle cast(const MapType &src, return_value_policy policy, handle parent) {
-        switch (policy) {
-            case return_value_policy::copy:
-                return eigen_array_cast<props>(src);
-            case return_value_policy::reference_internal:
-                return eigen_array_cast<props>(src, parent, is_eigen_mutable_map<MapType>::value);
-            case return_value_policy::reference:
-            case return_value_policy::automatic:
-            case return_value_policy::automatic_reference:
-                return eigen_array_cast<props>(src, none(), is_eigen_mutable_map<MapType>::value);
-            default:
-                // move, take_ownership don't make any sense for a ref/map:
-                pybind11_fail("Invalid return_value_policy for Eigen Map/Ref/Block type");
-        }
-    }
-
-    static constexpr auto name = props::descriptor;
-
-    // Explicitly delete these: support python -> C++ conversion on these (i.e. these can be return
-    // types but not bound arguments).  We still provide them (with an explicitly delete) so that
-    // you end up here if you try anyway.
-    bool load(handle, bool) = delete;
-    operator MapType() = delete;
-    template <typename>
-    using cast_op_type = MapType;
-};
-
-// We can return any map-like object (but can only load Refs, specialized next):
-template <typename Type>
-struct type_caster<Type, enable_if_t<is_eigen_dense_map<Type>::value>> : eigen_map_caster<Type> {};
-
-// Loader for Ref<...> arguments.  See the documentation for info on how to make this work without
-// copying (it requires some extra effort in many cases).
-template <typename PlainObjectType, typename StrideType>
-struct type_caster<
-    Eigen::Ref<PlainObjectType, 0, StrideType>,
-    enable_if_t<is_eigen_dense_map<Eigen::Ref<PlainObjectType, 0, StrideType>>::value>>
-    : public eigen_map_caster<Eigen::Ref<PlainObjectType, 0, StrideType>> {
-private:
-    using Type = Eigen::Ref<PlainObjectType, 0, StrideType>;
-    using props = EigenProps<Type>;
-    using Scalar = typename props::Scalar;
-    using MapType = Eigen::Map<PlainObjectType, 0, StrideType>;
-    using Array
-        = array_t<Scalar,
-                  array::forcecast
-                      | ((props::row_major ? props::inner_stride : props::outer_stride) == 1
-                             ? array::c_style
-                         : (props::row_major ? props::outer_stride : props::inner_stride) == 1
-                             ? array::f_style
-                             : 0)>;
-    static constexpr bool need_writeable = is_eigen_mutable_map<Type>::value;
-    // Delay construction (these have no default constructor)
-    std::unique_ptr<MapType> map;
-    std::unique_ptr<Type> ref;
-    // Our array.  When possible, this is just a numpy array pointing to the source data, but
-    // sometimes we can't avoid copying (e.g. input is not a numpy array at all, has an
-    // incompatible layout, or is an array of a type that needs to be converted).  Using a numpy
-    // temporary (rather than an Eigen temporary) saves an extra copy when we need both type
-    // conversion and storage order conversion.  (Note that we refuse to use this temporary copy
-    // when loading an argument for a Ref<M> with M non-const, i.e. a read-write reference).
-    Array copy_or_ref;
-
-public:
-    bool load(handle src, bool convert) {
-        // First check whether what we have is already an array of the right type.  If not, we
-        // can't avoid a copy (because the copy is also going to do type conversion).
-        bool need_copy = !isinstance<Array>(src);
-
-        EigenConformable<props::row_major> fits;
-        if (!need_copy) {
-            // We don't need a converting copy, but we also need to check whether the strides are
-            // compatible with the Ref's stride requirements
-            auto aref = reinterpret_borrow<Array>(src);
-
-            if (aref && (!need_writeable || aref.writeable())) {
-                fits = props::conformable(aref);
-                if (!fits) {
-                    return false; // Incompatible dimensions
-                }
-                if (!fits.template stride_compatible<props>()) {
-                    need_copy = true;
-                } else {
-                    copy_or_ref = std::move(aref);
-                }
-            } else {
-                need_copy = true;
-            }
-        }
-
-        if (need_copy) {
-            // We need to copy: If we need a mutable reference, or we're not supposed to convert
-            // (either because we're in the no-convert overload pass, or because we're explicitly
-            // instructed not to copy (via `py::arg().noconvert()`) we have to fail loading.
-            if (!convert || need_writeable) {
-                return false;
-            }
-
-            Array copy = Array::ensure(src);
-            if (!copy) {
-                return false;
-            }
-            fits = props::conformable(copy);
-            if (!fits || !fits.template stride_compatible<props>()) {
-                return false;
-            }
-            copy_or_ref = std::move(copy);
-            loader_life_support::add_patient(copy_or_ref);
-        }
-
-        ref.reset();
-        map.reset(new MapType(data(copy_or_ref),
-                              fits.rows,
-                              fits.cols,
-                              make_stride(fits.stride.outer(), fits.stride.inner())));
-        ref.reset(new Type(*map));
-
-        return true;
-    }
-
-    // NOLINTNEXTLINE(google-explicit-constructor)
-    operator Type *() { return ref.get(); }
-    // NOLINTNEXTLINE(google-explicit-constructor)
-    operator Type &() { return *ref; }
-    template <typename _T>
-    using cast_op_type = pybind11::detail::cast_op_type<_T>;
-
-private:
-    template <typename T = Type, enable_if_t<is_eigen_mutable_map<T>::value, int> = 0>
-    Scalar *data(Array &a) {
-        return a.mutable_data();
-    }
-
-    template <typename T = Type, enable_if_t<!is_eigen_mutable_map<T>::value, int> = 0>
-    const Scalar *data(Array &a) {
-        return a.data();
-    }
-
-    // Attempt to figure out a constructor of `Stride` that will work.
-    // If both strides are fixed, use a default constructor:
-    template <typename S>
-    using stride_ctor_default = bool_constant<S::InnerStrideAtCompileTime != Eigen::Dynamic
-                                              && S::OuterStrideAtCompileTime != Eigen::Dynamic
-                                              && std::is_default_constructible<S>::value>;
-    // Otherwise, if there is a two-index constructor, assume it is (outer,inner) like
-    // Eigen::Stride, and use it:
-    template <typename S>
-    using stride_ctor_dual
-        = bool_constant<!stride_ctor_default<S>::value
-                        && std::is_constructible<S, EigenIndex, EigenIndex>::value>;
-    // Otherwise, if there is a one-index constructor, and just one of the strides is dynamic, use
-    // it (passing whichever stride is dynamic).
-    template <typename S>
-    using stride_ctor_outer
-        = bool_constant<!any_of<stride_ctor_default<S>, stride_ctor_dual<S>>::value
-                        && S::OuterStrideAtCompileTime == Eigen::Dynamic
-                        && S::InnerStrideAtCompileTime != Eigen::Dynamic
-                        && std::is_constructible<S, EigenIndex>::value>;
-    template <typename S>
-    using stride_ctor_inner
-        = bool_constant<!any_of<stride_ctor_default<S>, stride_ctor_dual<S>>::value
-                        && S::InnerStrideAtCompileTime == Eigen::Dynamic
-                        && S::OuterStrideAtCompileTime != Eigen::Dynamic
-                        && std::is_constructible<S, EigenIndex>::value>;
-
-    template <typename S = StrideType, enable_if_t<stride_ctor_default<S>::value, int> = 0>
-    static S make_stride(EigenIndex, EigenIndex) {
-        return S();
-    }
-    template <typename S = StrideType, enable_if_t<stride_ctor_dual<S>::value, int> = 0>
-    static S make_stride(EigenIndex outer, EigenIndex inner) {
-        return S(outer, inner);
-    }
-    template <typename S = StrideType, enable_if_t<stride_ctor_outer<S>::value, int> = 0>
-    static S make_stride(EigenIndex outer, EigenIndex) {
-        return S(outer);
-    }
-    template <typename S = StrideType, enable_if_t<stride_ctor_inner<S>::value, int> = 0>
-    static S make_stride(EigenIndex, EigenIndex inner) {
-        return S(inner);
-    }
-};
-
-// type_caster for special matrix types (e.g. DiagonalMatrix), which are EigenBase, but not
-// EigenDense (i.e. they don't have a data(), at least not with the usual matrix layout).
-// load() is not supported, but we can cast them into the python domain by first copying to a
-// regular Eigen::Matrix, then casting that.
-template <typename Type>
-struct type_caster<Type, enable_if_t<is_eigen_other<Type>::value>> {
-protected:
-    using Matrix
-        = Eigen::Matrix<typename Type::Scalar, Type::RowsAtCompileTime, Type::ColsAtCompileTime>;
-    using props = EigenProps<Matrix>;
-
-public:
-    static handle cast(const Type &src, return_value_policy /* policy */, handle /* parent */) {
-        handle h = eigen_encapsulate<props>(new Matrix(src));
-        return h;
-    }
-    static handle cast(const Type *src, return_value_policy policy, handle parent) {
-        return cast(*src, policy, parent);
-    }
-
-    static constexpr auto name = props::descriptor;
-
-    // Explicitly delete these: support python -> C++ conversion on these (i.e. these can be return
-    // types but not bound arguments).  We still provide them (with an explicitly delete) so that
-    // you end up here if you try anyway.
-    bool load(handle, bool) = delete;
-    operator Type() = delete;
-    template <typename>
-    using cast_op_type = Type;
-};
-
-template <typename Type>
-struct type_caster<Type, enable_if_t<is_eigen_sparse<Type>::value>> {
-    using Scalar = typename Type::Scalar;
-    using StorageIndex = remove_reference_t<decltype(*std::declval<Type>().outerIndexPtr())>;
-    using Index = typename Type::Index;
-    static constexpr bool rowMajor = Type::IsRowMajor;
-
-    bool load(handle src, bool) {
-        if (!src) {
-            return false;
-        }
-
-        auto obj = reinterpret_borrow<object>(src);
-        object sparse_module = module_::import("scipy.sparse");
-        object matrix_type = sparse_module.attr(rowMajor ? "csr_matrix" : "csc_matrix");
-
-        if (!type::handle_of(obj).is(matrix_type)) {
-            try {
-                obj = matrix_type(obj);
-            } catch (const error_already_set &) {
-                return false;
-            }
-        }
-
-        auto values = array_t<Scalar>((object) obj.attr("data"));
-        auto innerIndices = array_t<StorageIndex>((object) obj.attr("indices"));
-        auto outerIndices = array_t<StorageIndex>((object) obj.attr("indptr"));
-        auto shape = pybind11::tuple((pybind11::object) obj.attr("shape"));
-        auto nnz = obj.attr("nnz").cast<Index>();
-
-        if (!values || !innerIndices || !outerIndices) {
-            return false;
-        }
-
-        value = EigenMapSparseMatrix<Scalar,
-                                     Type::Flags &(Eigen::RowMajor | Eigen::ColMajor),
-                                     StorageIndex>(shape[0].cast<Index>(),
-                                                   shape[1].cast<Index>(),
-                                                   std::move(nnz),
-                                                   outerIndices.mutable_data(),
-                                                   innerIndices.mutable_data(),
-                                                   values.mutable_data());
-
-        return true;
-    }
-
-    static handle cast(const Type &src, return_value_policy /* policy */, handle /* parent */) {
-        const_cast<Type &>(src).makeCompressed();
-
-        object matrix_type
-            = module_::import("scipy.sparse").attr(rowMajor ? "csr_matrix" : "csc_matrix");
-
-        array data(src.nonZeros(), src.valuePtr());
-        array outerIndices((rowMajor ? src.rows() : src.cols()) + 1, src.outerIndexPtr());
-        array innerIndices(src.nonZeros(), src.innerIndexPtr());
-
-        return matrix_type(std::make_tuple(
-                               std::move(data), std::move(innerIndices), std::move(outerIndices)),
-                           std::make_pair(src.rows(), src.cols()))
-            .release();
-    }
-
-    PYBIND11_TYPE_CASTER(Type,
-                         const_name<(Type::IsRowMajor) != 0>("scipy.sparse.csr_matrix[",
-                                                             "scipy.sparse.csc_matrix[")
-                             + npy_format_descriptor<Scalar>::name + const_name("]"));
-};
-
-PYBIND11_NAMESPACE_END(detail)
-PYBIND11_NAMESPACE_END(PYBIND11_NAMESPACE)
+#include "eigen/matrix.h"
diff --git a/include/pybind11/eigen/common.h b/include/pybind11/eigen/common.h
new file mode 100644
index 0000000000..24f56d1584
--- /dev/null
+++ b/include/pybind11/eigen/common.h
@@ -0,0 +1,9 @@
+// Copyright (c) 2023 The pybind Community.
+
+#pragma once
+
+// Common message for `static_assert()`s, which are useful to easily
+// preempt much less obvious errors.
+#define PYBIND11_EIGEN_MESSAGE_POINTER_TYPES_ARE_NOT_SUPPORTED                                    \
+    "Pointer types (in particular `PyObject *`) are not supported as scalar types for Eigen "     \
+    "types."
diff --git a/include/pybind11/eigen/matrix.h b/include/pybind11/eigen/matrix.h
new file mode 100644
index 0000000000..8d4342f81b
--- /dev/null
+++ b/include/pybind11/eigen/matrix.h
@@ -0,0 +1,714 @@
+/*
+    pybind11/eigen/matrix.h: Transparent conversion for dense and sparse Eigen matrices
+
+    Copyright (c) 2016 Wenzel Jakob <wenzel.jakob@epfl.ch>
+
+    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 "../numpy.h"
+#include "common.h"
+
+/* HINT: To suppress warnings originating from the Eigen headers, use -isystem.
+   See also:
+       https://stackoverflow.com/questions/2579576/i-dir-vs-isystem-dir
+       https://stackoverflow.com/questions/1741816/isystem-for-ms-visual-studio-c-compiler
+*/
+PYBIND11_WARNING_PUSH
+PYBIND11_WARNING_DISABLE_MSVC(5054) // https://github.com/pybind/pybind11/pull/3741
+//       C5054: operator '&': deprecated between enumerations of different types
+#if defined(__MINGW32__)
+PYBIND11_WARNING_DISABLE_GCC("-Wmaybe-uninitialized")
+#endif
+
+#include <Eigen/Core>
+#include <Eigen/SparseCore>
+
+PYBIND11_WARNING_POP
+
+// Eigen prior to 3.2.7 doesn't have proper move constructors--but worse, some classes get implicit
+// move constructors that break things.  We could detect this an explicitly copy, but an extra copy
+// of matrices seems highly undesirable.
+static_assert(EIGEN_VERSION_AT_LEAST(3, 2, 7),
+              "Eigen matrix support in pybind11 requires Eigen >= 3.2.7");
+
+PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)
+
+PYBIND11_WARNING_DISABLE_MSVC(4127)
+
+// Provide a convenience alias for easier pass-by-ref usage with fully dynamic strides:
+using EigenDStride = Eigen::Stride<Eigen::Dynamic, Eigen::Dynamic>;
+template <typename MatrixType>
+using EigenDRef = Eigen::Ref<MatrixType, 0, EigenDStride>;
+template <typename MatrixType>
+using EigenDMap = Eigen::Map<MatrixType, 0, EigenDStride>;
+
+PYBIND11_NAMESPACE_BEGIN(detail)
+
+#if EIGEN_VERSION_AT_LEAST(3, 3, 0)
+using EigenIndex = Eigen::Index;
+template <typename Scalar, int Flags, typename StorageIndex>
+using EigenMapSparseMatrix = Eigen::Map<Eigen::SparseMatrix<Scalar, Flags, StorageIndex>>;
+#else
+using EigenIndex = EIGEN_DEFAULT_DENSE_INDEX_TYPE;
+template <typename Scalar, int Flags, typename StorageIndex>
+using EigenMapSparseMatrix = Eigen::MappedSparseMatrix<Scalar, Flags, StorageIndex>;
+#endif
+
+// Matches Eigen::Map, Eigen::Ref, blocks, etc:
+template <typename T>
+using is_eigen_dense_map = all_of<is_template_base_of<Eigen::DenseBase, T>,
+                                  std::is_base_of<Eigen::MapBase<T, Eigen::ReadOnlyAccessors>, T>>;
+template <typename T>
+using is_eigen_mutable_map = std::is_base_of<Eigen::MapBase<T, Eigen::WriteAccessors>, T>;
+template <typename T>
+using is_eigen_dense_plain
+    = all_of<negation<is_eigen_dense_map<T>>, is_template_base_of<Eigen::PlainObjectBase, T>>;
+template <typename T>
+using is_eigen_sparse = is_template_base_of<Eigen::SparseMatrixBase, T>;
+// Test for objects inheriting from EigenBase<Derived> that aren't captured by the above.  This
+// basically covers anything that can be assigned to a dense matrix but that don't have a typical
+// matrix data layout that can be copied from their .data().  For example, DiagonalMatrix and
+// SelfAdjointView fall into this category.
+template <typename T>
+using is_eigen_other
+    = all_of<is_template_base_of<Eigen::EigenBase, T>,
+             negation<any_of<is_eigen_dense_map<T>, is_eigen_dense_plain<T>, is_eigen_sparse<T>>>>;
+
+// Captures numpy/eigen conformability status (returned by EigenProps::conformable()):
+template <bool EigenRowMajor>
+struct EigenConformable {
+    bool conformable = false;
+    EigenIndex rows = 0, cols = 0;
+    EigenDStride stride{0, 0};    // Only valid if negativestrides is false!
+    bool negativestrides = false; // If true, do not use stride!
+
+    // NOLINTNEXTLINE(google-explicit-constructor)
+    EigenConformable(bool fits = false) : conformable{fits} {}
+    // Matrix type:
+    EigenConformable(EigenIndex r, EigenIndex c, EigenIndex rstride, EigenIndex cstride)
+        : conformable{true}, rows{r}, cols{c},
+          // TODO: when Eigen bug #747 is fixed, remove the tests for non-negativity.
+          // http://eigen.tuxfamily.org/bz/show_bug.cgi?id=747
+          stride{EigenRowMajor ? (rstride > 0 ? rstride : 0)
+                               : (cstride > 0 ? cstride : 0) /* outer stride */,
+                 EigenRowMajor ? (cstride > 0 ? cstride : 0)
+                               : (rstride > 0 ? rstride : 0) /* inner stride */},
+          negativestrides{rstride < 0 || cstride < 0} {}
+    // Vector type:
+    EigenConformable(EigenIndex r, EigenIndex c, EigenIndex stride)
+        : EigenConformable(r, c, r == 1 ? c * stride : stride, c == 1 ? r : r * stride) {}
+
+    template <typename props>
+    bool stride_compatible() const {
+        // To have compatible strides, we need (on both dimensions) one of fully dynamic strides,
+        // matching strides, or a dimension size of 1 (in which case the stride value is
+        // irrelevant). Alternatively, if any dimension size is 0, the strides are not relevant
+        // (and numpy ≥ 1.23 sets the strides to 0 in that case, so we need to check explicitly).
+        if (negativestrides) {
+            return false;
+        }
+        if (rows == 0 || cols == 0) {
+            return true;
+        }
+        return (props::inner_stride == Eigen::Dynamic || props::inner_stride == stride.inner()
+                || (EigenRowMajor ? cols : rows) == 1)
+               && (props::outer_stride == Eigen::Dynamic || props::outer_stride == stride.outer()
+                   || (EigenRowMajor ? rows : cols) == 1);
+    }
+    // NOLINTNEXTLINE(google-explicit-constructor)
+    operator bool() const { return conformable; }
+};
+
+template <typename Type>
+struct eigen_extract_stride {
+    using type = Type;
+};
+template <typename PlainObjectType, int MapOptions, typename StrideType>
+struct eigen_extract_stride<Eigen::Map<PlainObjectType, MapOptions, StrideType>> {
+    using type = StrideType;
+};
+template <typename PlainObjectType, int Options, typename StrideType>
+struct eigen_extract_stride<Eigen::Ref<PlainObjectType, Options, StrideType>> {
+    using type = StrideType;
+};
+
+// Helper struct for extracting information from an Eigen type
+template <typename Type_>
+struct EigenProps {
+    using Type = Type_;
+    using Scalar = typename Type::Scalar;
+    using StrideType = typename eigen_extract_stride<Type>::type;
+    static constexpr EigenIndex rows = Type::RowsAtCompileTime, cols = Type::ColsAtCompileTime,
+                                size = Type::SizeAtCompileTime;
+    static constexpr bool row_major = Type::IsRowMajor,
+                          vector
+                          = Type::IsVectorAtCompileTime, // At least one dimension has fixed size 1
+        fixed_rows = rows != Eigen::Dynamic, fixed_cols = cols != Eigen::Dynamic,
+                          fixed = size != Eigen::Dynamic, // Fully-fixed size
+        dynamic = !fixed_rows && !fixed_cols;             // Fully-dynamic size
+
+    template <EigenIndex i, EigenIndex ifzero>
+    using if_zero = std::integral_constant<EigenIndex, i == 0 ? ifzero : i>;
+    static constexpr EigenIndex inner_stride
+        = if_zero<StrideType::InnerStrideAtCompileTime, 1>::value,
+        outer_stride = if_zero < StrideType::OuterStrideAtCompileTime,
+        vector      ? size
+        : row_major ? cols
+                    : rows > ::value;
+    static constexpr bool dynamic_stride
+        = inner_stride == Eigen::Dynamic && outer_stride == Eigen::Dynamic;
+    static constexpr bool requires_row_major
+        = !dynamic_stride && !vector && (row_major ? inner_stride : outer_stride) == 1;
+    static constexpr bool requires_col_major
+        = !dynamic_stride && !vector && (row_major ? outer_stride : inner_stride) == 1;
+
+    // Takes an input array and determines whether we can make it fit into the Eigen type.  If
+    // the array is a vector, we attempt to fit it into either an Eigen 1xN or Nx1 vector
+    // (preferring the latter if it will fit in either, i.e. for a fully dynamic matrix type).
+    static EigenConformable<row_major> conformable(const array &a) {
+        const auto dims = a.ndim();
+        if (dims < 1 || dims > 2) {
+            return false;
+        }
+
+        if (dims == 2) { // Matrix type: require exact match (or dynamic)
+
+            EigenIndex np_rows = a.shape(0), np_cols = a.shape(1),
+                       np_rstride = a.strides(0) / static_cast<ssize_t>(sizeof(Scalar)),
+                       np_cstride = a.strides(1) / static_cast<ssize_t>(sizeof(Scalar));
+            if ((fixed_rows && np_rows != rows) || (fixed_cols && np_cols != cols)) {
+                return false;
+            }
+
+            return {np_rows, np_cols, np_rstride, np_cstride};
+        }
+
+        // Otherwise we're storing an n-vector.  Only one of the strides will be used, but
+        // whichever is used, we want the (single) numpy stride value.
+        const EigenIndex n = a.shape(0),
+                         stride = a.strides(0) / static_cast<ssize_t>(sizeof(Scalar));
+
+        if (vector) { // Eigen type is a compile-time vector
+            if (fixed && size != n) {
+                return false; // Vector size mismatch
+            }
+            return {rows == 1 ? 1 : n, cols == 1 ? 1 : n, stride};
+        }
+        if (fixed) {
+            // The type has a fixed size, but is not a vector: abort
+            return false;
+        }
+        if (fixed_cols) {
+            // Since this isn't a vector, cols must be != 1.  We allow this only if it exactly
+            // equals the number of elements (rows is Dynamic, and so 1 row is allowed).
+            if (cols != n) {
+                return false;
+            }
+            return {1, n, stride};
+        } // Otherwise it's either fully dynamic, or column dynamic; both become a column vector
+        if (fixed_rows && rows != n) {
+            return false;
+        }
+        return {n, 1, stride};
+    }
+
+    static constexpr bool show_writeable
+        = is_eigen_dense_map<Type>::value && is_eigen_mutable_map<Type>::value;
+    static constexpr bool show_order = is_eigen_dense_map<Type>::value;
+    static constexpr bool show_c_contiguous = show_order && requires_row_major;
+    static constexpr bool show_f_contiguous
+        = !show_c_contiguous && show_order && requires_col_major;
+
+    static constexpr auto descriptor
+        = const_name("numpy.ndarray[") + npy_format_descriptor<Scalar>::name + const_name("[")
+          + const_name<fixed_rows>(const_name<(size_t) rows>(), const_name("m")) + const_name(", ")
+          + const_name<fixed_cols>(const_name<(size_t) cols>(), const_name("n")) + const_name("]")
+          +
+          // For a reference type (e.g. Ref<MatrixXd>) we have other constraints that might need to
+          // be satisfied: writeable=True (for a mutable reference), and, depending on the map's
+          // stride options, possibly f_contiguous or c_contiguous.  We include them in the
+          // descriptor output to provide some hint as to why a TypeError is occurring (otherwise
+          // it can be confusing to see that a function accepts a 'numpy.ndarray[float64[3,2]]' and
+          // an error message that you *gave* a numpy.ndarray of the right type and dimensions.
+          const_name<show_writeable>(", flags.writeable", "")
+          + const_name<show_c_contiguous>(", flags.c_contiguous", "")
+          + const_name<show_f_contiguous>(", flags.f_contiguous", "") + const_name("]");
+};
+
+// Casts an Eigen type to numpy array.  If given a base, the numpy array references the src data,
+// otherwise it'll make a copy.  writeable lets you turn off the writeable flag for the array.
+template <typename props>
+handle
+eigen_array_cast(typename props::Type const &src, handle base = handle(), bool writeable = true) {
+    constexpr ssize_t elem_size = sizeof(typename props::Scalar);
+    array a;
+    if (props::vector) {
+        a = array({src.size()}, {elem_size * src.innerStride()}, src.data(), base);
+    } else {
+        a = array({src.rows(), src.cols()},
+                  {elem_size * src.rowStride(), elem_size * src.colStride()},
+                  src.data(),
+                  base);
+    }
+
+    if (!writeable) {
+        array_proxy(a.ptr())->flags &= ~detail::npy_api::NPY_ARRAY_WRITEABLE_;
+    }
+
+    return a.release();
+}
+
+// Takes an lvalue ref to some Eigen type and a (python) base object, creating a numpy array that
+// reference the Eigen object's data with `base` as the python-registered base class (if omitted,
+// the base will be set to None, and lifetime management is up to the caller).  The numpy array is
+// non-writeable if the given type is const.
+template <typename props, typename Type>
+handle eigen_ref_array(Type &src, handle parent = none()) {
+    // none here is to get past array's should-we-copy detection, which currently always
+    // copies when there is no base.  Setting the base to None should be harmless.
+    return eigen_array_cast<props>(src, parent, !std::is_const<Type>::value);
+}
+
+// Takes a pointer to some dense, plain Eigen type, builds a capsule around it, then returns a
+// numpy array that references the encapsulated data with a python-side reference to the capsule to
+// tie its destruction to that of any dependent python objects.  Const-ness is determined by
+// whether or not the Type of the pointer given is const.
+template <typename props, typename Type, typename = enable_if_t<is_eigen_dense_plain<Type>::value>>
+handle eigen_encapsulate(Type *src) {
+    capsule base(src, [](void *o) { delete static_cast<Type *>(o); });
+    return eigen_ref_array<props>(*src, base);
+}
+
+// Type caster for regular, dense matrix types (e.g. MatrixXd), but not maps/refs/etc. of dense
+// types.
+template <typename Type>
+struct type_caster<Type, enable_if_t<is_eigen_dense_plain<Type>::value>> {
+    using Scalar = typename Type::Scalar;
+    static_assert(!std::is_pointer<Scalar>::value,
+                  PYBIND11_EIGEN_MESSAGE_POINTER_TYPES_ARE_NOT_SUPPORTED);
+    using props = EigenProps<Type>;
+
+    bool load(handle src, bool convert) {
+        // If we're in no-convert mode, only load if given an array of the correct type
+        if (!convert && !isinstance<array_t<Scalar>>(src)) {
+            return false;
+        }
+
+        // Coerce into an array, but don't do type conversion yet; the copy below handles it.
+        auto buf = array::ensure(src);
+
+        if (!buf) {
+            return false;
+        }
+
+        auto dims = buf.ndim();
+        if (dims < 1 || dims > 2) {
+            return false;
+        }
+
+        auto fits = props::conformable(buf);
+        if (!fits) {
+            return false;
+        }
+
+        // Allocate the new type, then build a numpy reference into it
+        value = Type(fits.rows, fits.cols);
+        auto ref = reinterpret_steal<array>(eigen_ref_array<props>(value));
+        if (dims == 1) {
+            ref = ref.squeeze();
+        } else if (ref.ndim() == 1) {
+            buf = buf.squeeze();
+        }
+
+        int result = detail::npy_api::get().PyArray_CopyInto_(ref.ptr(), buf.ptr());
+
+        if (result < 0) { // Copy failed!
+            PyErr_Clear();
+            return false;
+        }
+
+        return true;
+    }
+
+private:
+    // Cast implementation
+    template <typename CType>
+    static handle cast_impl(CType *src, return_value_policy policy, handle parent) {
+        switch (policy) {
+            case return_value_policy::take_ownership:
+            case return_value_policy::automatic:
+                return eigen_encapsulate<props>(src);
+            case return_value_policy::move:
+                return eigen_encapsulate<props>(new CType(std::move(*src)));
+            case return_value_policy::copy:
+                return eigen_array_cast<props>(*src);
+            case return_value_policy::reference:
+            case return_value_policy::automatic_reference:
+                return eigen_ref_array<props>(*src);
+            case return_value_policy::reference_internal:
+                return eigen_ref_array<props>(*src, parent);
+            default:
+                throw cast_error("unhandled return_value_policy: should not happen!");
+        };
+    }
+
+public:
+    // Normal returned non-reference, non-const value:
+    static handle cast(Type &&src, return_value_policy /* policy */, handle parent) {
+        return cast_impl(&src, return_value_policy::move, parent);
+    }
+    // If you return a non-reference const, we mark the numpy array readonly:
+    static handle cast(const Type &&src, return_value_policy /* policy */, handle parent) {
+        return cast_impl(&src, return_value_policy::move, parent);
+    }
+    // lvalue reference return; default (automatic) becomes copy
+    static handle cast(Type &src, return_value_policy policy, handle parent) {
+        if (policy == return_value_policy::automatic
+            || policy == return_value_policy::automatic_reference) {
+            policy = return_value_policy::copy;
+        }
+        return cast_impl(&src, policy, parent);
+    }
+    // const lvalue reference return; default (automatic) becomes copy
+    static handle cast(const Type &src, return_value_policy policy, handle parent) {
+        if (policy == return_value_policy::automatic
+            || policy == return_value_policy::automatic_reference) {
+            policy = return_value_policy::copy;
+        }
+        return cast(&src, policy, parent);
+    }
+    // non-const pointer return
+    static handle cast(Type *src, return_value_policy policy, handle parent) {
+        return cast_impl(src, policy, parent);
+    }
+    // const pointer return
+    static handle cast(const Type *src, return_value_policy policy, handle parent) {
+        return cast_impl(src, policy, parent);
+    }
+
+    static constexpr auto name = props::descriptor;
+
+    // NOLINTNEXTLINE(google-explicit-constructor)
+    operator Type *() { return &value; }
+    // NOLINTNEXTLINE(google-explicit-constructor)
+    operator Type &() { return value; }
+    // NOLINTNEXTLINE(google-explicit-constructor)
+    operator Type &&() && { return std::move(value); }
+    template <typename T>
+    using cast_op_type = movable_cast_op_type<T>;
+
+private:
+    Type value;
+};
+
+// Base class for casting reference/map/block/etc. objects back to python.
+template <typename MapType>
+struct eigen_map_caster {
+    static_assert(!std::is_pointer<typename MapType::Scalar>::value,
+                  PYBIND11_EIGEN_MESSAGE_POINTER_TYPES_ARE_NOT_SUPPORTED);
+
+private:
+    using props = EigenProps<MapType>;
+
+public:
+    // Directly referencing a ref/map's data is a bit dangerous (whatever the map/ref points to has
+    // to stay around), but we'll allow it under the assumption that you know what you're doing
+    // (and have an appropriate keep_alive in place).  We return a numpy array pointing directly at
+    // the ref's data (The numpy array ends up read-only if the ref was to a const matrix type.)
+    // Note that this means you need to ensure you don't destroy the object in some other way (e.g.
+    // with an appropriate keep_alive, or with a reference to a statically allocated matrix).
+    static handle cast(const MapType &src, return_value_policy policy, handle parent) {
+        switch (policy) {
+            case return_value_policy::copy:
+                return eigen_array_cast<props>(src);
+            case return_value_policy::reference_internal:
+                return eigen_array_cast<props>(src, parent, is_eigen_mutable_map<MapType>::value);
+            case return_value_policy::reference:
+            case return_value_policy::automatic:
+            case return_value_policy::automatic_reference:
+                return eigen_array_cast<props>(src, none(), is_eigen_mutable_map<MapType>::value);
+            default:
+                // move, take_ownership don't make any sense for a ref/map:
+                pybind11_fail("Invalid return_value_policy for Eigen Map/Ref/Block type");
+        }
+    }
+
+    static constexpr auto name = props::descriptor;
+
+    // Explicitly delete these: support python -> C++ conversion on these (i.e. these can be return
+    // types but not bound arguments).  We still provide them (with an explicitly delete) so that
+    // you end up here if you try anyway.
+    bool load(handle, bool) = delete;
+    operator MapType() = delete;
+    template <typename>
+    using cast_op_type = MapType;
+};
+
+// We can return any map-like object (but can only load Refs, specialized next):
+template <typename Type>
+struct type_caster<Type, enable_if_t<is_eigen_dense_map<Type>::value>> : eigen_map_caster<Type> {};
+
+// Loader for Ref<...> arguments.  See the documentation for info on how to make this work without
+// copying (it requires some extra effort in many cases).
+template <typename PlainObjectType, typename StrideType>
+struct type_caster<
+    Eigen::Ref<PlainObjectType, 0, StrideType>,
+    enable_if_t<is_eigen_dense_map<Eigen::Ref<PlainObjectType, 0, StrideType>>::value>>
+    : public eigen_map_caster<Eigen::Ref<PlainObjectType, 0, StrideType>> {
+private:
+    using Type = Eigen::Ref<PlainObjectType, 0, StrideType>;
+    using props = EigenProps<Type>;
+    using Scalar = typename props::Scalar;
+    static_assert(!std::is_pointer<Scalar>::value,
+                  PYBIND11_EIGEN_MESSAGE_POINTER_TYPES_ARE_NOT_SUPPORTED);
+    using MapType = Eigen::Map<PlainObjectType, 0, StrideType>;
+    using Array
+        = array_t<Scalar,
+                  array::forcecast
+                      | ((props::row_major ? props::inner_stride : props::outer_stride) == 1
+                             ? array::c_style
+                         : (props::row_major ? props::outer_stride : props::inner_stride) == 1
+                             ? array::f_style
+                             : 0)>;
+    static constexpr bool need_writeable = is_eigen_mutable_map<Type>::value;
+    // Delay construction (these have no default constructor)
+    std::unique_ptr<MapType> map;
+    std::unique_ptr<Type> ref;
+    // Our array.  When possible, this is just a numpy array pointing to the source data, but
+    // sometimes we can't avoid copying (e.g. input is not a numpy array at all, has an
+    // incompatible layout, or is an array of a type that needs to be converted).  Using a numpy
+    // temporary (rather than an Eigen temporary) saves an extra copy when we need both type
+    // conversion and storage order conversion.  (Note that we refuse to use this temporary copy
+    // when loading an argument for a Ref<M> with M non-const, i.e. a read-write reference).
+    Array copy_or_ref;
+
+public:
+    bool load(handle src, bool convert) {
+        // First check whether what we have is already an array of the right type.  If not, we
+        // can't avoid a copy (because the copy is also going to do type conversion).
+        bool need_copy = !isinstance<Array>(src);
+
+        EigenConformable<props::row_major> fits;
+        if (!need_copy) {
+            // We don't need a converting copy, but we also need to check whether the strides are
+            // compatible with the Ref's stride requirements
+            auto aref = reinterpret_borrow<Array>(src);
+
+            if (aref && (!need_writeable || aref.writeable())) {
+                fits = props::conformable(aref);
+                if (!fits) {
+                    return false; // Incompatible dimensions
+                }
+                if (!fits.template stride_compatible<props>()) {
+                    need_copy = true;
+                } else {
+                    copy_or_ref = std::move(aref);
+                }
+            } else {
+                need_copy = true;
+            }
+        }
+
+        if (need_copy) {
+            // We need to copy: If we need a mutable reference, or we're not supposed to convert
+            // (either because we're in the no-convert overload pass, or because we're explicitly
+            // instructed not to copy (via `py::arg().noconvert()`) we have to fail loading.
+            if (!convert || need_writeable) {
+                return false;
+            }
+
+            Array copy = Array::ensure(src);
+            if (!copy) {
+                return false;
+            }
+            fits = props::conformable(copy);
+            if (!fits || !fits.template stride_compatible<props>()) {
+                return false;
+            }
+            copy_or_ref = std::move(copy);
+            loader_life_support::add_patient(copy_or_ref);
+        }
+
+        ref.reset();
+        map.reset(new MapType(data(copy_or_ref),
+                              fits.rows,
+                              fits.cols,
+                              make_stride(fits.stride.outer(), fits.stride.inner())));
+        ref.reset(new Type(*map));
+
+        return true;
+    }
+
+    // NOLINTNEXTLINE(google-explicit-constructor)
+    operator Type *() { return ref.get(); }
+    // NOLINTNEXTLINE(google-explicit-constructor)
+    operator Type &() { return *ref; }
+    template <typename _T>
+    using cast_op_type = pybind11::detail::cast_op_type<_T>;
+
+private:
+    template <typename T = Type, enable_if_t<is_eigen_mutable_map<T>::value, int> = 0>
+    Scalar *data(Array &a) {
+        return a.mutable_data();
+    }
+
+    template <typename T = Type, enable_if_t<!is_eigen_mutable_map<T>::value, int> = 0>
+    const Scalar *data(Array &a) {
+        return a.data();
+    }
+
+    // Attempt to figure out a constructor of `Stride` that will work.
+    // If both strides are fixed, use a default constructor:
+    template <typename S>
+    using stride_ctor_default = bool_constant<S::InnerStrideAtCompileTime != Eigen::Dynamic
+                                              && S::OuterStrideAtCompileTime != Eigen::Dynamic
+                                              && std::is_default_constructible<S>::value>;
+    // Otherwise, if there is a two-index constructor, assume it is (outer,inner) like
+    // Eigen::Stride, and use it:
+    template <typename S>
+    using stride_ctor_dual
+        = bool_constant<!stride_ctor_default<S>::value
+                        && std::is_constructible<S, EigenIndex, EigenIndex>::value>;
+    // Otherwise, if there is a one-index constructor, and just one of the strides is dynamic, use
+    // it (passing whichever stride is dynamic).
+    template <typename S>
+    using stride_ctor_outer
+        = bool_constant<!any_of<stride_ctor_default<S>, stride_ctor_dual<S>>::value
+                        && S::OuterStrideAtCompileTime == Eigen::Dynamic
+                        && S::InnerStrideAtCompileTime != Eigen::Dynamic
+                        && std::is_constructible<S, EigenIndex>::value>;
+    template <typename S>
+    using stride_ctor_inner
+        = bool_constant<!any_of<stride_ctor_default<S>, stride_ctor_dual<S>>::value
+                        && S::InnerStrideAtCompileTime == Eigen::Dynamic
+                        && S::OuterStrideAtCompileTime != Eigen::Dynamic
+                        && std::is_constructible<S, EigenIndex>::value>;
+
+    template <typename S = StrideType, enable_if_t<stride_ctor_default<S>::value, int> = 0>
+    static S make_stride(EigenIndex, EigenIndex) {
+        return S();
+    }
+    template <typename S = StrideType, enable_if_t<stride_ctor_dual<S>::value, int> = 0>
+    static S make_stride(EigenIndex outer, EigenIndex inner) {
+        return S(outer, inner);
+    }
+    template <typename S = StrideType, enable_if_t<stride_ctor_outer<S>::value, int> = 0>
+    static S make_stride(EigenIndex outer, EigenIndex) {
+        return S(outer);
+    }
+    template <typename S = StrideType, enable_if_t<stride_ctor_inner<S>::value, int> = 0>
+    static S make_stride(EigenIndex, EigenIndex inner) {
+        return S(inner);
+    }
+};
+
+// type_caster for special matrix types (e.g. DiagonalMatrix), which are EigenBase, but not
+// EigenDense (i.e. they don't have a data(), at least not with the usual matrix layout).
+// load() is not supported, but we can cast them into the python domain by first copying to a
+// regular Eigen::Matrix, then casting that.
+template <typename Type>
+struct type_caster<Type, enable_if_t<is_eigen_other<Type>::value>> {
+    static_assert(!std::is_pointer<typename Type::Scalar>::value,
+                  PYBIND11_EIGEN_MESSAGE_POINTER_TYPES_ARE_NOT_SUPPORTED);
+
+protected:
+    using Matrix
+        = Eigen::Matrix<typename Type::Scalar, Type::RowsAtCompileTime, Type::ColsAtCompileTime>;
+    using props = EigenProps<Matrix>;
+
+public:
+    static handle cast(const Type &src, return_value_policy /* policy */, handle /* parent */) {
+        handle h = eigen_encapsulate<props>(new Matrix(src));
+        return h;
+    }
+    static handle cast(const Type *src, return_value_policy policy, handle parent) {
+        return cast(*src, policy, parent);
+    }
+
+    static constexpr auto name = props::descriptor;
+
+    // Explicitly delete these: support python -> C++ conversion on these (i.e. these can be return
+    // types but not bound arguments).  We still provide them (with an explicitly delete) so that
+    // you end up here if you try anyway.
+    bool load(handle, bool) = delete;
+    operator Type() = delete;
+    template <typename>
+    using cast_op_type = Type;
+};
+
+template <typename Type>
+struct type_caster<Type, enable_if_t<is_eigen_sparse<Type>::value>> {
+    using Scalar = typename Type::Scalar;
+    static_assert(!std::is_pointer<Scalar>::value,
+                  PYBIND11_EIGEN_MESSAGE_POINTER_TYPES_ARE_NOT_SUPPORTED);
+    using StorageIndex = remove_reference_t<decltype(*std::declval<Type>().outerIndexPtr())>;
+    using Index = typename Type::Index;
+    static constexpr bool rowMajor = Type::IsRowMajor;
+
+    bool load(handle src, bool) {
+        if (!src) {
+            return false;
+        }
+
+        auto obj = reinterpret_borrow<object>(src);
+        object sparse_module = module_::import("scipy.sparse");
+        object matrix_type = sparse_module.attr(rowMajor ? "csr_matrix" : "csc_matrix");
+
+        if (!type::handle_of(obj).is(matrix_type)) {
+            try {
+                obj = matrix_type(obj);
+            } catch (const error_already_set &) {
+                return false;
+            }
+        }
+
+        auto values = array_t<Scalar>((object) obj.attr("data"));
+        auto innerIndices = array_t<StorageIndex>((object) obj.attr("indices"));
+        auto outerIndices = array_t<StorageIndex>((object) obj.attr("indptr"));
+        auto shape = pybind11::tuple((pybind11::object) obj.attr("shape"));
+        auto nnz = obj.attr("nnz").cast<Index>();
+
+        if (!values || !innerIndices || !outerIndices) {
+            return false;
+        }
+
+        value = EigenMapSparseMatrix<Scalar,
+                                     Type::Flags &(Eigen::RowMajor | Eigen::ColMajor),
+                                     StorageIndex>(shape[0].cast<Index>(),
+                                                   shape[1].cast<Index>(),
+                                                   std::move(nnz),
+                                                   outerIndices.mutable_data(),
+                                                   innerIndices.mutable_data(),
+                                                   values.mutable_data());
+
+        return true;
+    }
+
+    static handle cast(const Type &src, return_value_policy /* policy */, handle /* parent */) {
+        const_cast<Type &>(src).makeCompressed();
+
+        object matrix_type
+            = module_::import("scipy.sparse").attr(rowMajor ? "csr_matrix" : "csc_matrix");
+
+        array data(src.nonZeros(), src.valuePtr());
+        array outerIndices((rowMajor ? src.rows() : src.cols()) + 1, src.outerIndexPtr());
+        array innerIndices(src.nonZeros(), src.innerIndexPtr());
+
+        return matrix_type(pybind11::make_tuple(
+                               std::move(data), std::move(innerIndices), std::move(outerIndices)),
+                           pybind11::make_tuple(src.rows(), src.cols()))
+            .release();
+    }
+
+    PYBIND11_TYPE_CASTER(Type,
+                         const_name<(Type::IsRowMajor) != 0>("scipy.sparse.csr_matrix[",
+                                                             "scipy.sparse.csc_matrix[")
+                             + npy_format_descriptor<Scalar>::name + const_name("]"));
+};
+
+PYBIND11_NAMESPACE_END(detail)
+PYBIND11_NAMESPACE_END(PYBIND11_NAMESPACE)
diff --git a/include/pybind11/eigen/tensor.h b/include/pybind11/eigen/tensor.h
new file mode 100644
index 0000000000..d4ed6c0ca8
--- /dev/null
+++ b/include/pybind11/eigen/tensor.h
@@ -0,0 +1,517 @@
+/*
+    pybind11/eigen/tensor.h: Transparent conversion for Eigen tensors
+
+    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 "../numpy.h"
+#include "common.h"
+
+#if defined(__GNUC__) && !defined(__clang__) && !defined(__INTEL_COMPILER)
+static_assert(__GNUC__ > 5, "Eigen Tensor support in pybind11 requires GCC > 5.0");
+#endif
+
+// Disable warnings for Eigen
+PYBIND11_WARNING_PUSH
+PYBIND11_WARNING_DISABLE_MSVC(4554)
+PYBIND11_WARNING_DISABLE_MSVC(4127)
+#if defined(__MINGW32__)
+PYBIND11_WARNING_DISABLE_GCC("-Wmaybe-uninitialized")
+#endif
+
+#include <unsupported/Eigen/CXX11/Tensor>
+
+PYBIND11_WARNING_POP
+
+static_assert(EIGEN_VERSION_AT_LEAST(3, 3, 0),
+              "Eigen Tensor support in pybind11 requires Eigen >= 3.3.0");
+
+PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)
+
+PYBIND11_WARNING_DISABLE_MSVC(4127)
+
+PYBIND11_NAMESPACE_BEGIN(detail)
+
+inline bool is_tensor_aligned(const void *data) {
+    return (reinterpret_cast<std::size_t>(data) % EIGEN_DEFAULT_ALIGN_BYTES) == 0;
+}
+
+template <typename T>
+constexpr int compute_array_flag_from_tensor() {
+    static_assert((static_cast<int>(T::Layout) == static_cast<int>(Eigen::RowMajor))
+                      || (static_cast<int>(T::Layout) == static_cast<int>(Eigen::ColMajor)),
+                  "Layout must be row or column major");
+    return (static_cast<int>(T::Layout) == static_cast<int>(Eigen::RowMajor)) ? array::c_style
+                                                                              : array::f_style;
+}
+
+template <typename T>
+struct eigen_tensor_helper {};
+
+template <typename Scalar_, int NumIndices_, int Options_, typename IndexType>
+struct eigen_tensor_helper<Eigen::Tensor<Scalar_, NumIndices_, Options_, IndexType>> {
+    using Type = Eigen::Tensor<Scalar_, NumIndices_, Options_, IndexType>;
+    using ValidType = void;
+
+    static Eigen::DSizes<typename Type::Index, Type::NumIndices> get_shape(const Type &f) {
+        return f.dimensions();
+    }
+
+    static constexpr bool
+    is_correct_shape(const Eigen::DSizes<typename Type::Index, Type::NumIndices> & /*shape*/) {
+        return true;
+    }
+
+    template <typename T>
+    struct helper {};
+
+    template <size_t... Is>
+    struct helper<index_sequence<Is...>> {
+        static constexpr auto value = ::pybind11::detail::concat(const_name(((void) Is, "?"))...);
+    };
+
+    static constexpr auto dimensions_descriptor
+        = helper<decltype(make_index_sequence<Type::NumIndices>())>::value;
+
+    template <typename... Args>
+    static Type *alloc(Args &&...args) {
+        return new Type(std::forward<Args>(args)...);
+    }
+
+    static void free(Type *tensor) { delete tensor; }
+};
+
+template <typename Scalar_, typename std::ptrdiff_t... Indices, int Options_, typename IndexType>
+struct eigen_tensor_helper<
+    Eigen::TensorFixedSize<Scalar_, Eigen::Sizes<Indices...>, Options_, IndexType>> {
+    using Type = Eigen::TensorFixedSize<Scalar_, Eigen::Sizes<Indices...>, Options_, IndexType>;
+    using ValidType = void;
+
+    static constexpr Eigen::DSizes<typename Type::Index, Type::NumIndices>
+    get_shape(const Type & /*f*/) {
+        return get_shape();
+    }
+
+    static constexpr Eigen::DSizes<typename Type::Index, Type::NumIndices> get_shape() {
+        return Eigen::DSizes<typename Type::Index, Type::NumIndices>(Indices...);
+    }
+
+    static bool
+    is_correct_shape(const Eigen::DSizes<typename Type::Index, Type::NumIndices> &shape) {
+        return get_shape() == shape;
+    }
+
+    static constexpr auto dimensions_descriptor
+        = ::pybind11::detail::concat(const_name<Indices>()...);
+
+    template <typename... Args>
+    static Type *alloc(Args &&...args) {
+        Eigen::aligned_allocator<Type> allocator;
+        return ::new (allocator.allocate(1)) Type(std::forward<Args>(args)...);
+    }
+
+    static void free(Type *tensor) {
+        Eigen::aligned_allocator<Type> allocator;
+        tensor->~Type();
+        allocator.deallocate(tensor, 1);
+    }
+};
+
+template <typename Type, bool ShowDetails, bool NeedsWriteable = false>
+struct get_tensor_descriptor {
+    static constexpr auto details
+        = const_name<NeedsWriteable>(", flags.writeable", "")
+          + const_name<static_cast<int>(Type::Layout) == 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
+          + const_name("]") + const_name<ShowDetails>(details, const_name("")) + const_name("]");
+};
+
+// When EIGEN_AVOID_STL_ARRAY is defined, Eigen::DSizes<T, 0> does not have the begin() member
+// function. Falling back to a simple loop works around this issue.
+//
+// We need to disable the type-limits warning for the inner loop when size = 0.
+
+PYBIND11_WARNING_PUSH
+PYBIND11_WARNING_DISABLE_GCC("-Wtype-limits")
+
+template <typename T, int size>
+std::vector<T> convert_dsizes_to_vector(const Eigen::DSizes<T, size> &arr) {
+    std::vector<T> result(size);
+
+    for (size_t i = 0; i < size; i++) {
+        result[i] = arr[i];
+    }
+
+    return result;
+}
+
+template <typename T, int size>
+Eigen::DSizes<T, size> get_shape_for_array(const array &arr) {
+    Eigen::DSizes<T, size> result;
+    const T *shape = arr.shape();
+    for (size_t i = 0; i < size; i++) {
+        result[i] = shape[i];
+    }
+
+    return result;
+}
+
+PYBIND11_WARNING_POP
+
+template <typename Type>
+struct type_caster<Type, typename eigen_tensor_helper<Type>::ValidType> {
+    static_assert(!std::is_pointer<typename Type::Scalar>::value,
+                  PYBIND11_EIGEN_MESSAGE_POINTER_TYPES_ARE_NOT_SUPPORTED);
+    using Helper = eigen_tensor_helper<Type>;
+    static constexpr auto temp_name = get_tensor_descriptor<Type, false>::value;
+    PYBIND11_TYPE_CASTER(Type, temp_name);
+
+    bool load(handle src, bool convert) {
+        if (!convert) {
+            if (!isinstance<array>(src)) {
+                return false;
+            }
+            array temp = array::ensure(src);
+            if (!temp) {
+                return false;
+            }
+
+            if (!temp.dtype().is(dtype::of<typename Type::Scalar>())) {
+                return false;
+            }
+        }
+
+        array_t<typename Type::Scalar, compute_array_flag_from_tensor<Type>()> arr(
+            reinterpret_borrow<object>(src));
+
+        if (arr.ndim() != Type::NumIndices) {
+            return false;
+        }
+        auto shape = get_shape_for_array<typename Type::Index, Type::NumIndices>(arr);
+
+        if (!Helper::is_correct_shape(shape)) {
+            return false;
+        }
+
+#if EIGEN_VERSION_AT_LEAST(3, 4, 0)
+        auto data_pointer = arr.data();
+#else
+        // Handle Eigen bug
+        auto data_pointer = const_cast<typename Type::Scalar *>(arr.data());
+#endif
+
+        if (is_tensor_aligned(arr.data())) {
+            value = Eigen::TensorMap<const Type, Eigen::Aligned>(data_pointer, shape);
+        } else {
+            value = Eigen::TensorMap<const Type>(data_pointer, shape);
+        }
+
+        return true;
+    }
+
+    static handle cast(Type &&src, return_value_policy policy, handle parent) {
+        if (policy == return_value_policy::reference
+            || policy == return_value_policy::reference_internal) {
+            pybind11_fail("Cannot use a reference return value policy for an rvalue");
+        }
+        return cast_impl(&src, return_value_policy::move, parent);
+    }
+
+    static handle cast(const Type &&src, return_value_policy policy, handle parent) {
+        if (policy == return_value_policy::reference
+            || policy == return_value_policy::reference_internal) {
+            pybind11_fail("Cannot use a reference return value policy for an rvalue");
+        }
+        return cast_impl(&src, return_value_policy::move, parent);
+    }
+
+    static handle cast(Type &src, return_value_policy policy, handle parent) {
+        if (policy == return_value_policy::automatic
+            || policy == return_value_policy::automatic_reference) {
+            policy = return_value_policy::copy;
+        }
+        return cast_impl(&src, policy, parent);
+    }
+
+    static handle cast(const Type &src, return_value_policy policy, handle parent) {
+        if (policy == return_value_policy::automatic
+            || policy == return_value_policy::automatic_reference) {
+            policy = return_value_policy::copy;
+        }
+        return cast(&src, policy, parent);
+    }
+
+    static handle cast(Type *src, return_value_policy policy, handle parent) {
+        if (policy == return_value_policy::automatic) {
+            policy = return_value_policy::take_ownership;
+        } else if (policy == return_value_policy::automatic_reference) {
+            policy = return_value_policy::reference;
+        }
+        return cast_impl(src, policy, parent);
+    }
+
+    static handle cast(const Type *src, return_value_policy policy, handle parent) {
+        if (policy == return_value_policy::automatic) {
+            policy = return_value_policy::take_ownership;
+        } else if (policy == return_value_policy::automatic_reference) {
+            policy = return_value_policy::reference;
+        }
+        return cast_impl(src, policy, parent);
+    }
+
+    template <typename C>
+    static handle cast_impl(C *src, return_value_policy policy, handle parent) {
+        object parent_object;
+        bool writeable = false;
+        switch (policy) {
+            case return_value_policy::move:
+                if (std::is_const<C>::value) {
+                    pybind11_fail("Cannot move from a constant reference");
+                }
+
+                src = Helper::alloc(std::move(*src));
+
+                parent_object
+                    = capsule(src, [](void *ptr) { Helper::free(reinterpret_cast<Type *>(ptr)); });
+                writeable = true;
+                break;
+
+            case return_value_policy::take_ownership:
+                if (std::is_const<C>::value) {
+                    // This cast is ugly, and might be UB in some cases, but we don't have an
+                    // alternative here as we must free that memory
+                    Helper::free(const_cast<Type *>(src));
+                    pybind11_fail("Cannot take ownership of a const reference");
+                }
+
+                parent_object
+                    = capsule(src, [](void *ptr) { Helper::free(reinterpret_cast<Type *>(ptr)); });
+                writeable = true;
+                break;
+
+            case return_value_policy::copy:
+                writeable = true;
+                break;
+
+            case return_value_policy::reference:
+                parent_object = none();
+                writeable = !std::is_const<C>::value;
+                break;
+
+            case return_value_policy::reference_internal:
+                // Default should do the right thing
+                if (!parent) {
+                    pybind11_fail("Cannot use reference internal when there is no parent");
+                }
+                parent_object = reinterpret_borrow<object>(parent);
+                writeable = !std::is_const<C>::value;
+                break;
+
+            default:
+                pybind11_fail("pybind11 bug in eigen.h, please file a bug report");
+        }
+
+        auto result = array_t<typename Type::Scalar, compute_array_flag_from_tensor<Type>()>(
+            convert_dsizes_to_vector(Helper::get_shape(*src)), src->data(), parent_object);
+
+        if (!writeable) {
+            array_proxy(result.ptr())->flags &= ~detail::npy_api::NPY_ARRAY_WRITEABLE_;
+        }
+
+        return result.release();
+    }
+};
+
+template <typename StoragePointerType,
+          bool needs_writeable,
+          enable_if_t<!needs_writeable, bool> = true>
+StoragePointerType get_array_data_for_type(array &arr) {
+#if EIGEN_VERSION_AT_LEAST(3, 4, 0)
+    return reinterpret_cast<StoragePointerType>(arr.data());
+#else
+    // Handle Eigen bug
+    return reinterpret_cast<StoragePointerType>(const_cast<void *>(arr.data()));
+#endif
+}
+
+template <typename StoragePointerType,
+          bool needs_writeable,
+          enable_if_t<needs_writeable, bool> = true>
+StoragePointerType get_array_data_for_type(array &arr) {
+    return reinterpret_cast<StoragePointerType>(arr.mutable_data());
+}
+
+template <typename T, typename = void>
+struct get_storage_pointer_type;
+
+template <typename MapType>
+struct get_storage_pointer_type<MapType, void_t<typename MapType::StoragePointerType>> {
+    using SPT = typename MapType::StoragePointerType;
+};
+
+template <typename MapType>
+struct get_storage_pointer_type<MapType, void_t<typename MapType::PointerArgType>> {
+    using SPT = typename MapType::PointerArgType;
+};
+
+template <typename Type, int Options>
+struct type_caster<Eigen::TensorMap<Type, Options>,
+                   typename eigen_tensor_helper<remove_cv_t<Type>>::ValidType> {
+    static_assert(!std::is_pointer<typename Type::Scalar>::value,
+                  PYBIND11_EIGEN_MESSAGE_POINTER_TYPES_ARE_NOT_SUPPORTED);
+    using MapType = Eigen::TensorMap<Type, Options>;
+    using Helper = eigen_tensor_helper<remove_cv_t<Type>>;
+
+    bool load(handle src, bool /*convert*/) {
+        // Note that we have a lot more checks here as we want to make sure to avoid copies
+        if (!isinstance<array>(src)) {
+            return false;
+        }
+        auto arr = reinterpret_borrow<array>(src);
+        if ((arr.flags() & compute_array_flag_from_tensor<Type>()) == 0) {
+            return false;
+        }
+
+        if (!arr.dtype().is(dtype::of<typename Type::Scalar>())) {
+            return false;
+        }
+
+        if (arr.ndim() != Type::NumIndices) {
+            return false;
+        }
+
+        constexpr bool is_aligned = (Options & Eigen::Aligned) != 0;
+
+        if (is_aligned && !is_tensor_aligned(arr.data())) {
+            return false;
+        }
+
+        auto shape = get_shape_for_array<typename Type::Index, Type::NumIndices>(arr);
+
+        if (!Helper::is_correct_shape(shape)) {
+            return false;
+        }
+
+        if (needs_writeable && !arr.writeable()) {
+            return false;
+        }
+
+        auto result = get_array_data_for_type<typename get_storage_pointer_type<MapType>::SPT,
+                                              needs_writeable>(arr);
+
+        value.reset(new MapType(std::move(result), std::move(shape)));
+
+        return true;
+    }
+
+    static handle cast(MapType &&src, return_value_policy policy, handle parent) {
+        return cast_impl(&src, policy, parent);
+    }
+
+    static handle cast(const MapType &&src, return_value_policy policy, handle parent) {
+        return cast_impl(&src, policy, parent);
+    }
+
+    static handle cast(MapType &src, return_value_policy policy, handle parent) {
+        if (policy == return_value_policy::automatic
+            || policy == return_value_policy::automatic_reference) {
+            policy = return_value_policy::copy;
+        }
+        return cast_impl(&src, policy, parent);
+    }
+
+    static handle cast(const MapType &src, return_value_policy policy, handle parent) {
+        if (policy == return_value_policy::automatic
+            || policy == return_value_policy::automatic_reference) {
+            policy = return_value_policy::copy;
+        }
+        return cast(&src, policy, parent);
+    }
+
+    static handle cast(MapType *src, return_value_policy policy, handle parent) {
+        if (policy == return_value_policy::automatic) {
+            policy = return_value_policy::take_ownership;
+        } else if (policy == return_value_policy::automatic_reference) {
+            policy = return_value_policy::reference;
+        }
+        return cast_impl(src, policy, parent);
+    }
+
+    static handle cast(const MapType *src, return_value_policy policy, handle parent) {
+        if (policy == return_value_policy::automatic) {
+            policy = return_value_policy::take_ownership;
+        } else if (policy == return_value_policy::automatic_reference) {
+            policy = return_value_policy::reference;
+        }
+        return cast_impl(src, policy, parent);
+    }
+
+    template <typename C>
+    static handle cast_impl(C *src, return_value_policy policy, handle parent) {
+        object parent_object;
+        constexpr bool writeable = !std::is_const<C>::value;
+        switch (policy) {
+            case return_value_policy::reference:
+                parent_object = none();
+                break;
+
+            case return_value_policy::reference_internal:
+                // Default should do the right thing
+                if (!parent) {
+                    pybind11_fail("Cannot use reference internal when there is no parent");
+                }
+                parent_object = reinterpret_borrow<object>(parent);
+                break;
+
+            case return_value_policy::take_ownership:
+                delete src;
+                // fallthrough
+            default:
+                // move, take_ownership don't make any sense for a ref/map:
+                pybind11_fail("Invalid return_value_policy for Eigen Map type, must be either "
+                              "reference or reference_internal");
+        }
+
+        auto result = array_t<typename Type::Scalar, compute_array_flag_from_tensor<Type>()>(
+            convert_dsizes_to_vector(Helper::get_shape(*src)),
+            src->data(),
+            std::move(parent_object));
+
+        if (!writeable) {
+            array_proxy(result.ptr())->flags &= ~detail::npy_api::NPY_ARRAY_WRITEABLE_;
+        }
+
+        return result.release();
+    }
+
+#if EIGEN_VERSION_AT_LEAST(3, 4, 0)
+
+    static constexpr bool needs_writeable = !std::is_const<typename std::remove_pointer<
+        typename get_storage_pointer_type<MapType>::SPT>::type>::value;
+#else
+    // Handle Eigen bug
+    static constexpr bool needs_writeable = !std::is_const<Type>::value;
+#endif
+
+protected:
+    // TODO: Move to std::optional once std::optional has more support
+    std::unique_ptr<MapType> value;
+
+public:
+    static constexpr auto name = get_tensor_descriptor<Type, true, needs_writeable>::value;
+    explicit operator MapType *() { return value.get(); }
+    explicit operator MapType &() { return *value; }
+    explicit operator MapType &&() && { return std::move(*value); }
+
+    template <typename T_>
+    using cast_op_type = ::pybind11::detail::movable_cast_op_type<T_>;
+};
+
+PYBIND11_NAMESPACE_END(detail)
+PYBIND11_NAMESPACE_END(PYBIND11_NAMESPACE)
diff --git a/include/pybind11/embed.h b/include/pybind11/embed.h
index 98cc3e4664..9d29eb8249 100644
--- a/include/pybind11/embed.h
+++ b/include/pybind11/embed.h
@@ -86,8 +86,24 @@ inline wchar_t *widen_chars(const char *safe_arg) {
     return widened_arg;
 }
 
-/// Python 2.x/3.x-compatible version of `PySys_SetArgv`
-inline void set_interpreter_argv(int argc, const char *const *argv, bool add_program_dir_to_path) {
+inline void precheck_interpreter() {
+    if (Py_IsInitialized() != 0) {
+        pybind11_fail("The interpreter is already running");
+    }
+}
+
+#if !defined(PYBIND11_PYCONFIG_SUPPORT_PY_VERSION_HEX)
+#    define PYBIND11_PYCONFIG_SUPPORT_PY_VERSION_HEX (0x03080000)
+#endif
+
+#if PY_VERSION_HEX < PYBIND11_PYCONFIG_SUPPORT_PY_VERSION_HEX
+inline void initialize_interpreter_pre_pyconfig(bool init_signal_handlers,
+                                                int argc,
+                                                const char *const *argv,
+                                                bool add_program_dir_to_path) {
+    detail::precheck_interpreter();
+    Py_InitializeEx(init_signal_handlers ? 1 : 0);
+
     // Before it was special-cased in python 3.8, passing an empty or null argv
     // caused a segfault, so we have to reimplement the special case ourselves.
     bool special_case = (argv == nullptr || argc <= 0);
@@ -101,10 +117,10 @@ inline void set_interpreter_argv(int argc, const char *const *argv, bool add_pro
     auto argv_size = static_cast<size_t>(argc);
     // SetArgv* on python 3 takes wchar_t, so we have to convert.
     std::unique_ptr<wchar_t *[]> widened_argv(new wchar_t *[argv_size]);
-    std::vector<std::unique_ptr<wchar_t[], wide_char_arg_deleter>> widened_argv_entries;
+    std::vector<std::unique_ptr<wchar_t[], detail::wide_char_arg_deleter>> widened_argv_entries;
     widened_argv_entries.reserve(argv_size);
     for (size_t ii = 0; ii < argv_size; ++ii) {
-        widened_argv_entries.emplace_back(widen_chars(safe_argv[ii]));
+        widened_argv_entries.emplace_back(detail::widen_chars(safe_argv[ii]));
         if (!widened_argv_entries.back()) {
             // A null here indicates a character-encoding failure or the python
             // interpreter out of memory. Give up.
@@ -114,11 +130,43 @@ inline void set_interpreter_argv(int argc, const char *const *argv, bool add_pro
     }
 
     auto *pysys_argv = widened_argv.get();
+
     PySys_SetArgvEx(argc, pysys_argv, static_cast<int>(add_program_dir_to_path));
 }
+#endif
 
 PYBIND11_NAMESPACE_END(detail)
 
+#if PY_VERSION_HEX >= PYBIND11_PYCONFIG_SUPPORT_PY_VERSION_HEX
+inline void initialize_interpreter(PyConfig *config,
+                                   int argc = 0,
+                                   const char *const *argv = nullptr,
+                                   bool add_program_dir_to_path = true) {
+    detail::precheck_interpreter();
+    PyStatus status = PyConfig_SetBytesArgv(config, argc, const_cast<char *const *>(argv));
+    if (PyStatus_Exception(status) != 0) {
+        // A failure here indicates a character-encoding failure or the python
+        // interpreter out of memory. Give up.
+        PyConfig_Clear(config);
+        throw std::runtime_error(PyStatus_IsError(status) != 0 ? status.err_msg
+                                                               : "Failed to prepare CPython");
+    }
+    status = Py_InitializeFromConfig(config);
+    if (PyStatus_Exception(status) != 0) {
+        PyConfig_Clear(config);
+        throw std::runtime_error(PyStatus_IsError(status) != 0 ? status.err_msg
+                                                               : "Failed to init CPython");
+    }
+    if (add_program_dir_to_path) {
+        PyRun_SimpleString("import sys, os.path; "
+                           "sys.path.insert(0, "
+                           "os.path.abspath(os.path.dirname(sys.argv[0])) "
+                           "if sys.argv and os.path.exists(sys.argv[0]) else '')");
+    }
+    PyConfig_Clear(config);
+}
+#endif
+
 /** \rst
     Initialize the Python interpreter. No other pybind11 or CPython API functions can be
     called before this is done; with the exception of `PYBIND11_EMBEDDED_MODULE`. The
@@ -142,13 +190,18 @@ inline void initialize_interpreter(bool init_signal_handlers = true,
                                    int argc = 0,
                                    const char *const *argv = nullptr,
                                    bool add_program_dir_to_path = true) {
-    if (Py_IsInitialized() != 0) {
-        pybind11_fail("The interpreter is already running");
-    }
-
-    Py_InitializeEx(init_signal_handlers ? 1 : 0);
-
-    detail::set_interpreter_argv(argc, argv, add_program_dir_to_path);
+#if PY_VERSION_HEX < PYBIND11_PYCONFIG_SUPPORT_PY_VERSION_HEX
+    detail::initialize_interpreter_pre_pyconfig(
+        init_signal_handlers, argc, argv, add_program_dir_to_path);
+#else
+    PyConfig config;
+    PyConfig_InitPythonConfig(&config);
+    // See PR #4473 for background
+    config.parse_argv = 0;
+
+    config.install_signal_handlers = init_signal_handlers ? 1 : 0;
+    initialize_interpreter(&config, argc, argv, add_program_dir_to_path);
+#endif
 }
 
 /** \rst
@@ -187,16 +240,14 @@ inline void initialize_interpreter(bool init_signal_handlers = true,
 
  \endrst */
 inline void finalize_interpreter() {
-    handle builtins(PyEval_GetBuiltins());
-    const char *id = PYBIND11_INTERNALS_ID;
-
     // Get the internals pointer (without creating it if it doesn't exist).  It's possible for the
     // internals to be created during Py_Finalize() (e.g. if a py::capsule calls `get_internals()`
     // during destruction), so we get the pointer-pointer here and check it after Py_Finalize().
     detail::internals **internals_ptr_ptr = detail::get_internals_pp();
-    // It could also be stashed in builtins, so look there too:
-    if (builtins.contains(id) && isinstance<capsule>(builtins[id])) {
-        internals_ptr_ptr = capsule(builtins[id]);
+    // It could also be stashed in state_dict, so look there too:
+    if (object internals_obj
+        = get_internals_obj_from_state_dict(detail::get_python_state_dict())) {
+        internals_ptr_ptr = detail::get_internals_pp_from_capsule(internals_obj);
     }
     // Local internals contains data managed by the current interpreter, so we must clear them to
     // avoid undefined behaviors when initializing another interpreter
@@ -235,6 +286,15 @@ class scoped_interpreter {
         initialize_interpreter(init_signal_handlers, argc, argv, add_program_dir_to_path);
     }
 
+#if PY_VERSION_HEX >= PYBIND11_PYCONFIG_SUPPORT_PY_VERSION_HEX
+    explicit scoped_interpreter(PyConfig *config,
+                                int argc = 0,
+                                const char *const *argv = nullptr,
+                                bool add_program_dir_to_path = true) {
+        initialize_interpreter(config, argc, argv, add_program_dir_to_path);
+    }
+#endif
+
     scoped_interpreter(const scoped_interpreter &) = delete;
     scoped_interpreter(scoped_interpreter &&other) noexcept { other.is_valid = false; }
     scoped_interpreter &operator=(const scoped_interpreter &) = delete;
diff --git a/include/pybind11/functional.h b/include/pybind11/functional.h
index 23e092e731..6856119cde 100644
--- a/include/pybind11/functional.h
+++ b/include/pybind11/functional.h
@@ -48,9 +48,16 @@ struct type_caster<std::function<Return(Args...)>> {
          */
         if (auto cfunc = func.cpp_function()) {
             auto *cfunc_self = PyCFunction_GET_SELF(cfunc.ptr());
-            if (isinstance<capsule>(cfunc_self)) {
+            if (cfunc_self == nullptr) {
+                PyErr_Clear();
+            } else if (isinstance<capsule>(cfunc_self)) {
                 auto c = reinterpret_borrow<capsule>(cfunc_self);
-                auto *rec = (function_record *) c;
+
+                function_record *rec = nullptr;
+                // Check that we can safely reinterpret the capsule into a function_record
+                if (detail::is_function_record_capsule(c)) {
+                    rec = c.get_pointer<function_record>();
+                }
 
                 while (rec != nullptr) {
                     if (rec->is_stateless
@@ -98,8 +105,8 @@ struct type_caster<std::function<Return(Args...)>> {
             explicit func_wrapper(func_handle &&hf) noexcept : hfunc(std::move(hf)) {}
             Return operator()(Args... args) const {
                 gil_scoped_acquire acq;
-                object retval(hfunc.f(std::forward<Args>(args)...));
-                return retval.template cast<Return>();
+                // casts the returned object as a rvalue to the return type
+                return hfunc.f(std::forward<Args>(args)...).template cast<Return>();
             }
         };
 
@@ -110,7 +117,7 @@ struct type_caster<std::function<Return(Args...)>> {
     template <typename Func>
     static handle cast(Func &&f_, return_value_policy policy, handle /* parent */) {
         if (!f_) {
-            return none().inc_ref();
+            return none().release();
         }
 
         auto result = f_.template target<function_type>();
@@ -121,7 +128,8 @@ struct type_caster<std::function<Return(Args...)>> {
     }
 
     PYBIND11_TYPE_CASTER(type,
-                         const_name("Callable[[") + concat(make_caster<Args>::name...)
+                         const_name("Callable[[")
+                             + ::pybind11::detail::concat(make_caster<Args>::name...)
                              + const_name("], ") + make_caster<retval_type>::name
                              + const_name("]"));
 };
diff --git a/include/pybind11/gil.h b/include/pybind11/gil.h
index 446f4620b2..6b0edaee4e 100644
--- a/include/pybind11/gil.h
+++ b/include/pybind11/gil.h
@@ -10,7 +10,12 @@
 #pragma once
 
 #include "detail/common.h"
-#include "detail/internals.h"
+
+#include <cassert>
+
+#if !defined(PYBIND11_SIMPLE_GIL_MANAGEMENT)
+#    include "detail/internals.h"
+#endif
 
 PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)
 
@@ -21,7 +26,7 @@ PyThreadState *get_thread_state_unchecked();
 
 PYBIND11_NAMESPACE_END(detail)
 
-#if defined(WITH_THREAD) && !defined(PYPY_VERSION)
+#if !defined(PYBIND11_SIMPLE_GIL_MANAGEMENT)
 
 /* The functions below essentially reproduce the PyGILState_* API using a RAII
  * pattern, but there are a few important differences:
@@ -62,7 +67,7 @@ class gil_scoped_acquire {
 
         if (!tstate) {
             tstate = PyThreadState_New(internals.istate);
-#    if !defined(NDEBUG)
+#    if defined(PYBIND11_DETAILED_ERROR_MESSAGES)
             if (!tstate) {
                 pybind11_fail("scoped_acquire: could not create thread state!");
             }
@@ -80,11 +85,14 @@ class gil_scoped_acquire {
         inc_ref();
     }
 
+    gil_scoped_acquire(const gil_scoped_acquire &) = delete;
+    gil_scoped_acquire &operator=(const gil_scoped_acquire &) = delete;
+
     void inc_ref() { ++tstate->gilstate_counter; }
 
     PYBIND11_NOINLINE void dec_ref() {
         --tstate->gilstate_counter;
-#    if !defined(NDEBUG)
+#    if defined(PYBIND11_DETAILED_ERROR_MESSAGES)
         if (detail::get_thread_state_unchecked() != tstate) {
             pybind11_fail("scoped_acquire::dec_ref(): thread state must be current!");
         }
@@ -93,7 +101,7 @@ class gil_scoped_acquire {
         }
 #    endif
         if (tstate->gilstate_counter == 0) {
-#    if !defined(NDEBUG)
+#    if defined(PYBIND11_DETAILED_ERROR_MESSAGES)
             if (!release) {
                 pybind11_fail("scoped_acquire::dec_ref(): internal error!");
             }
@@ -129,7 +137,9 @@ class gil_scoped_acquire {
 
 class gil_scoped_release {
 public:
+    // PRECONDITION: The GIL must be held when this constructor is called.
     explicit gil_scoped_release(bool disassoc = false) : disassoc(disassoc) {
+        assert(PyGILState_Check());
         // `get_internals()` must be called here unconditionally in order to initialize
         // `internals.tstate` for subsequent `gil_scoped_acquire` calls. Otherwise, an
         // initialization race could occur as multiple threads try `gil_scoped_acquire`.
@@ -144,6 +154,9 @@ class gil_scoped_release {
         }
     }
 
+    gil_scoped_release(const gil_scoped_release &) = delete;
+    gil_scoped_release &operator=(const gil_scoped_release &) = delete;
+
     /// This method will disable the PyThreadState_DeleteCurrent call and the
     /// GIL won't be acquired. This method should be used if the interpreter
     /// could be shutting down when this is called, as thread deletion is not
@@ -172,12 +185,16 @@ class gil_scoped_release {
     bool disassoc;
     bool active = true;
 };
-#elif defined(PYPY_VERSION)
+
+#else // PYBIND11_SIMPLE_GIL_MANAGEMENT
+
 class gil_scoped_acquire {
     PyGILState_STATE state;
 
 public:
-    gil_scoped_acquire() { state = PyGILState_Ensure(); }
+    gil_scoped_acquire() : state{PyGILState_Ensure()} {}
+    gil_scoped_acquire(const gil_scoped_acquire &) = delete;
+    gil_scoped_acquire &operator=(const gil_scoped_acquire &) = delete;
     ~gil_scoped_acquire() { PyGILState_Release(state); }
     void disarm() {}
 };
@@ -186,17 +203,17 @@ class gil_scoped_release {
     PyThreadState *state;
 
 public:
-    gil_scoped_release() { state = PyEval_SaveThread(); }
+    // PRECONDITION: The GIL must be held when this constructor is called.
+    gil_scoped_release() {
+        assert(PyGILState_Check());
+        state = PyEval_SaveThread();
+    }
+    gil_scoped_release(const gil_scoped_release &) = delete;
+    gil_scoped_release &operator=(const gil_scoped_release &) = delete;
     ~gil_scoped_release() { PyEval_RestoreThread(state); }
     void disarm() {}
 };
-#else
-class gil_scoped_acquire {
-    void disarm() {}
-};
-class gil_scoped_release {
-    void disarm() {}
-};
-#endif
+
+#endif // PYBIND11_SIMPLE_GIL_MANAGEMENT
 
 PYBIND11_NAMESPACE_END(PYBIND11_NAMESPACE)
diff --git a/include/pybind11/gil_safe_call_once.h b/include/pybind11/gil_safe_call_once.h
new file mode 100644
index 0000000000..5f9e1b03c6
--- /dev/null
+++ b/include/pybind11/gil_safe_call_once.h
@@ -0,0 +1,100 @@
+// Copyright (c) 2023 The pybind Community.
+
+#pragma once
+
+#include "detail/common.h"
+#include "gil.h"
+
+#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
+//
+//   static auto imported_obj = py::module_::import("module_name"); // BAD, DO NOT USE!
+//
+// which has two serious issues:
+//
+//     1. Py_DECREF() calls potentially after the Python interpreter was finalized already, and
+//     2. deadlocks in multi-threaded processes (because of missing lock ordering).
+//
+// The following alternative avoids both problems:
+//
+//   PYBIND11_CONSTINIT static py::gil_safe_call_once_and_store<py::object> storage;
+//   auto &imported_obj = storage // Do NOT make this `static`!
+//       .call_once_and_store_result([]() {
+//           return py::module_::import("module_name");
+//       })
+//       .get_stored();
+//
+// The parameter of `call_once_and_store_result()` must be callable. It can make
+// CPython API calls, and in particular, it can temporarily release the GIL.
+//
+// `T` can be any C++ type, it does not have to involve CPython API types.
+//
+// The behavior with regard to signals, e.g. `SIGINT` (`KeyboardInterrupt`),
+// is not ideal. If the main thread is the one to actually run the `Callable`,
+// then a `KeyboardInterrupt` will interrupt it if it is running normal Python
+// code. The situation is different if a non-main thread runs the
+// `Callable`, and then the main thread starts waiting for it to complete:
+// a `KeyboardInterrupt` will not interrupt the non-main thread, but it will
+// get processed only when it is the main thread's turn again and it is running
+// normal Python code. However, this will be unnoticeable for quick call-once
+// functions, which is usually the case.
+template <typename T>
+class gil_safe_call_once_and_store {
+public:
+    // PRECONDITION: The GIL must be held when `call_once_and_store_result()` is called.
+    template <typename Callable>
+    gil_safe_call_once_and_store &call_once_and_store_result(Callable &&fn) {
+        if (!is_initialized_) { // This read is guarded by the GIL.
+            // Multiple threads may enter here, because the GIL is released in the next line and
+            // CPython API calls in the `fn()` call below may release and reacquire the GIL.
+            gil_scoped_release gil_rel; // Needed to establish lock ordering.
+            std::call_once(once_flag_, [&] {
+                // Only one thread will ever enter here.
+                gil_scoped_acquire gil_acq;
+                ::new (storage_) T(fn()); // fn may release, but will reacquire, the GIL.
+                is_initialized_ = true;   // This write is guarded by the GIL.
+            });
+            // All threads will observe `is_initialized_` as true here.
+        }
+        // Intentionally not returning `T &` to ensure the calling code is self-documenting.
+        return *this;
+    }
+
+    // This must only be called after `call_once_and_store_result()` was called.
+    T &get_stored() {
+        assert(is_initialized_);
+        PYBIND11_WARNING_PUSH
+#if !defined(__clang__) && defined(__GNUC__) && __GNUC__ < 5
+        // Needed for gcc 4.8.5
+        PYBIND11_WARNING_DISABLE_GCC("-Wstrict-aliasing")
+#endif
+        return *reinterpret_cast<T *>(storage_);
+        PYBIND11_WARNING_POP
+    }
+
+    constexpr gil_safe_call_once_and_store() = default;
+    PYBIND11_DTOR_CONSTEXPR ~gil_safe_call_once_and_store() = default;
+
+private:
+    alignas(T) char storage_[sizeof(T)] = {};
+    std::once_flag once_flag_ = {};
+#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.
+};
+
+PYBIND11_NAMESPACE_END(PYBIND11_NAMESPACE)
diff --git a/include/pybind11/iostream.h b/include/pybind11/iostream.h
index 12e1f19f0c..1878089e31 100644
--- a/include/pybind11/iostream.h
+++ b/include/pybind11/iostream.h
@@ -100,7 +100,7 @@ class pythonbuf : public std::streambuf {
 
             if (size > remainder) {
                 str line(pbase(), size - remainder);
-                pywrite(line);
+                pywrite(std::move(line));
                 pyflush();
             }
 
diff --git a/include/pybind11/numpy.h b/include/pybind11/numpy.h
index 7624c9fbf4..05ef3918b1 100644
--- a/include/pybind11/numpy.h
+++ b/include/pybind11/numpy.h
@@ -10,7 +10,10 @@
 #pragma once
 
 #include "pybind11.h"
+#include "detail/common.h"
 #include "complex.h"
+#include "gil_safe_call_once.h"
+#include "pytypes.h"
 
 #include <algorithm>
 #include <array>
@@ -26,20 +29,33 @@
 #include <utility>
 #include <vector>
 
+#if defined(PYBIND11_NUMPY_1_ONLY) && !defined(PYBIND11_INTERNAL_NUMPY_1_ONLY_DETECTED)
+#    error PYBIND11_NUMPY_1_ONLY must be defined before any pybind11 header is included.
+#endif
+
 /* This will be true on all flat address space platforms and allows us to reduce the
    whole npy_intp / ssize_t / Py_intptr_t business down to just ssize_t for all size
    and dimension types (e.g. shape, strides, indexing), instead of inflicting this
-   upon the library user. */
+   upon the library user.
+   Note that NumPy 2 now uses ssize_t for `npy_intp` to simplify this. */
 static_assert(sizeof(::pybind11::ssize_t) == sizeof(Py_intptr_t), "ssize_t != Py_intptr_t");
 static_assert(std::is_signed<Py_intptr_t>::value, "Py_intptr_t must be signed");
 // We now can reinterpret_cast between py::ssize_t and Py_intptr_t (MSVC + PyPy cares)
 
 PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)
 
+PYBIND11_WARNING_DISABLE_MSVC(4127)
+
+class dtype; // Forward declaration
 class array; // Forward declaration
 
 PYBIND11_NAMESPACE_BEGIN(detail)
 
+template <>
+struct handle_type_name<dtype> {
+    static constexpr auto name = const_name("numpy.dtype");
+};
+
 template <>
 struct handle_type_name<array> {
     static constexpr auto name = const_name("numpy.ndarray");
@@ -48,7 +64,8 @@ struct handle_type_name<array> {
 template <typename type, typename SFINAE = void>
 struct npy_format_descriptor;
 
-struct PyArrayDescr_Proxy {
+/* NumPy 1 proxy (always includes legacy fields) */
+struct PyArrayDescr1_Proxy {
     PyObject_HEAD
     PyObject *typeobj;
     char kind;
@@ -63,6 +80,43 @@ struct PyArrayDescr_Proxy {
     PyObject *names;
 };
 
+#ifndef PYBIND11_NUMPY_1_ONLY
+struct PyArrayDescr_Proxy {
+    PyObject_HEAD
+    PyObject *typeobj;
+    char kind;
+    char type;
+    char byteorder;
+    char _former_flags;
+    int type_num;
+    /* Additional fields are NumPy version specific. */
+};
+#else
+/* NumPy 1.x only, we can expose all fields */
+using PyArrayDescr_Proxy = PyArrayDescr1_Proxy;
+#endif
+
+/* NumPy 2 proxy, including legacy fields */
+struct PyArrayDescr2_Proxy {
+    PyObject_HEAD
+    PyObject *typeobj;
+    char kind;
+    char type;
+    char byteorder;
+    char _former_flags;
+    int type_num;
+    std::uint64_t flags;
+    ssize_t elsize;
+    ssize_t alignment;
+    PyObject *metadata;
+    Py_hash_t hash;
+    void *reserved_null[2];
+    /* The following fields only exist if 0 <= type_num < 2056 */
+    char *subarray;
+    PyObject *fields;
+    PyObject *names;
+};
+
 struct PyArray_Proxy {
     PyObject_HEAD
     char *data;
@@ -118,6 +172,28 @@ inline numpy_internals &get_numpy_internals() {
     return *ptr;
 }
 
+PYBIND11_NOINLINE module_ import_numpy_core_submodule(const char *submodule_name) {
+    module_ numpy = module_::import("numpy");
+    str version_string = numpy.attr("__version__");
+
+    module_ numpy_lib = module_::import("numpy.lib");
+    object numpy_version = numpy_lib.attr("NumpyVersion")(version_string);
+    int major_version = numpy_version.attr("major").cast<int>();
+
+#ifdef PYBIND11_NUMPY_1_ONLY
+    if (major_version >= 2) {
+        throw std::runtime_error(
+            "This extension was built with PYBIND11_NUMPY_1_ONLY defined, "
+            "but NumPy 2 is used in this process. For NumPy2 compatibility, "
+            "this extension needs to be rebuilt without the PYBIND11_NUMPY_1_ONLY define.");
+    }
+#endif
+    /* `numpy.core` was renamed to `numpy._core` in NumPy 2.0 as it officially
+        became a private module. */
+    std::string numpy_core_path = major_version >= 2 ? "numpy._core" : "numpy.core";
+    return module_::import((numpy_core_path + "." + submodule_name).c_str());
+}
+
 template <typename T>
 struct same_size {
     template <typename U>
@@ -184,14 +260,16 @@ struct npy_api {
             NPY_ULONG_, NPY_ULONGLONG_, NPY_UINT_),
     };
 
+    unsigned int PyArray_RUNTIME_VERSION_;
+
     struct PyArray_Dims {
         Py_intptr_t *ptr;
         int len;
     };
 
     static npy_api &get() {
-        static npy_api api = lookup();
-        return api;
+        PYBIND11_CONSTINIT static gil_safe_call_once_and_store<npy_api> storage;
+        return storage.call_once_and_store_result(lookup).get_stored();
     }
 
     bool PyArray_Check_(PyObject *obj) const {
@@ -222,6 +300,7 @@ struct npy_api {
     PyObject *(*PyArray_FromAny_)(PyObject *, PyObject *, int, int, int, PyObject *);
     int (*PyArray_DescrConverter_)(PyObject *, PyObject **);
     bool (*PyArray_EquivTypes_)(PyObject *, PyObject *);
+#ifdef PYBIND11_NUMPY_1_ONLY
     int (*PyArray_GetArrayParamsFromObject_)(PyObject *,
                                              PyObject *,
                                              unsigned char,
@@ -230,6 +309,7 @@ struct npy_api {
                                              Py_intptr_t *,
                                              PyObject **,
                                              PyObject *);
+#endif
     PyObject *(*PyArray_Squeeze_)(PyObject *);
     // Unused. Not removed because that affects ABI of the class.
     int (*PyArray_SetBaseObject_)(PyObject *, PyObject *);
@@ -247,7 +327,8 @@ struct npy_api {
         API_PyArray_DescrFromScalar = 57,
         API_PyArray_FromAny = 69,
         API_PyArray_Resize = 80,
-        API_PyArray_CopyInto = 82,
+        // CopyInto was slot 82 and 50 was effectively an alias. NumPy 2 removed 82.
+        API_PyArray_CopyInto = 50,
         API_PyArray_NewCopy = 85,
         API_PyArray_NewFromDescr = 94,
         API_PyArray_DescrNewFromType = 96,
@@ -256,18 +337,25 @@ struct npy_api {
         API_PyArray_View = 137,
         API_PyArray_DescrConverter = 174,
         API_PyArray_EquivTypes = 182,
+#ifdef PYBIND11_NUMPY_1_ONLY
         API_PyArray_GetArrayParamsFromObject = 278,
+#endif
         API_PyArray_SetBaseObject = 282
     };
 
     static npy_api lookup() {
-        module_ m = module_::import("numpy.core.multiarray");
+        module_ m = detail::import_numpy_core_submodule("multiarray");
         auto c = m.attr("_ARRAY_API");
-        void **api_ptr = (void **) PyCapsule_GetPointer(c.ptr(), NULL);
+        void **api_ptr = (void **) PyCapsule_GetPointer(c.ptr(), nullptr);
+        if (api_ptr == nullptr) {
+            raise_from(PyExc_SystemError, "FAILURE obtaining numpy _ARRAY_API pointer.");
+            throw error_already_set();
+        }
         npy_api api;
 #define DECL_NPY_API(Func) api.Func##_ = (decltype(api.Func##_)) api_ptr[API_##Func];
         DECL_NPY_API(PyArray_GetNDArrayCFeatureVersion);
-        if (api.PyArray_GetNDArrayCFeatureVersion_() < 0x7) {
+        api.PyArray_RUNTIME_VERSION_ = api.PyArray_GetNDArrayCFeatureVersion_();
+        if (api.PyArray_RUNTIME_VERSION_ < 0x7) {
             pybind11_fail("pybind11 numpy support requires numpy >= 1.7.0");
         }
         DECL_NPY_API(PyArray_Type);
@@ -286,7 +374,9 @@ struct npy_api {
         DECL_NPY_API(PyArray_View);
         DECL_NPY_API(PyArray_DescrConverter);
         DECL_NPY_API(PyArray_EquivTypes);
+#ifdef PYBIND11_NUMPY_1_ONLY
         DECL_NPY_API(PyArray_GetArrayParamsFromObject);
+#endif
         DECL_NPY_API(PyArray_SetBaseObject);
 
 #undef DECL_NPY_API
@@ -308,6 +398,14 @@ inline const PyArrayDescr_Proxy *array_descriptor_proxy(const PyObject *ptr) {
     return reinterpret_cast<const PyArrayDescr_Proxy *>(ptr);
 }
 
+inline const PyArrayDescr1_Proxy *array_descriptor1_proxy(const PyObject *ptr) {
+    return reinterpret_cast<const PyArrayDescr1_Proxy *>(ptr);
+}
+
+inline const PyArrayDescr2_Proxy *array_descriptor2_proxy(const PyObject *ptr) {
+    return reinterpret_cast<const PyArrayDescr2_Proxy *>(ptr);
+}
+
 inline bool check_flags(const void *ptr, int flag) {
     return (flag == (array_proxy(ptr)->flags & flag));
 }
@@ -348,7 +446,7 @@ struct array_info<std::array<T, N>> {
     }
 
     static constexpr auto extents = const_name<array_info<T>::is_array>(
-        concat(const_name<N>(), array_info<T>::extents), const_name<N>());
+        ::pybind11::detail::concat(const_name<N>(), array_info<T>::extents), const_name<N>());
 };
 // For numpy we have special handling for arrays of characters, so we don't include
 // the size in the array extents.
@@ -537,21 +635,21 @@ PYBIND11_NAMESPACE_END(detail)
 
 class dtype : public object {
 public:
-    PYBIND11_OBJECT_DEFAULT(dtype, object, detail::npy_api::get().PyArrayDescr_Check_);
+    PYBIND11_OBJECT_DEFAULT(dtype, object, detail::npy_api::get().PyArrayDescr_Check_)
 
     explicit dtype(const buffer_info &info) {
-        dtype descr(_dtype_from_pep3118()(PYBIND11_STR_TYPE(info.format)));
+        dtype descr(_dtype_from_pep3118()(pybind11::str(info.format)));
         // If info.itemsize == 0, use the value calculated from the format string
         m_ptr = descr.strip_padding(info.itemsize != 0 ? info.itemsize : descr.itemsize())
                     .release()
                     .ptr();
     }
 
-    explicit dtype(const std::string &format) {
-        m_ptr = from_args(pybind11::str(format)).release().ptr();
-    }
+    explicit dtype(const pybind11::str &format) : dtype(from_args(format)) {}
 
-    explicit dtype(const char *format) : dtype(std::string(format)) {}
+    explicit dtype(const std::string &format) : dtype(pybind11::str(format)) {}
+
+    explicit dtype(const char *format) : dtype(pybind11::str(format)) {}
 
     dtype(list names, list formats, list offsets, ssize_t itemsize) {
         dict args;
@@ -559,11 +657,20 @@ class dtype : public object {
         args["formats"] = std::move(formats);
         args["offsets"] = std::move(offsets);
         args["itemsize"] = pybind11::int_(itemsize);
-        m_ptr = from_args(std::move(args)).release().ptr();
+        m_ptr = from_args(args).release().ptr();
+    }
+
+    /// Return dtype for the given typenum (one of the NPY_TYPES).
+    /// https://numpy.org/devdocs/reference/c-api/array.html#c.PyArray_DescrFromType
+    explicit dtype(int typenum)
+        : object(detail::npy_api::get().PyArray_DescrFromType_(typenum), stolen_t{}) {
+        if (m_ptr == nullptr) {
+            throw error_already_set();
+        }
     }
 
     /// This is essentially the same as calling numpy.dtype(args) in Python.
-    static dtype from_args(object args) {
+    static dtype from_args(const object &args) {
         PyObject *ptr = nullptr;
         if ((detail::npy_api::get().PyArray_DescrConverter_(args.ptr(), &ptr) == 0) || !ptr) {
             throw error_already_set();
@@ -578,10 +685,32 @@ class dtype : public object {
     }
 
     /// Size of the data type in bytes.
+#ifdef PYBIND11_NUMPY_1_ONLY
     ssize_t itemsize() const { return detail::array_descriptor_proxy(m_ptr)->elsize; }
+#else
+    ssize_t itemsize() const {
+        if (detail::npy_api::get().PyArray_RUNTIME_VERSION_ < 0x12) {
+            return detail::array_descriptor1_proxy(m_ptr)->elsize;
+        }
+        return detail::array_descriptor2_proxy(m_ptr)->elsize;
+    }
+#endif
 
     /// Returns true for structured data types.
+#ifdef PYBIND11_NUMPY_1_ONLY
     bool has_fields() const { return detail::array_descriptor_proxy(m_ptr)->names != nullptr; }
+#else
+    bool has_fields() const {
+        if (detail::npy_api::get().PyArray_RUNTIME_VERSION_ < 0x12) {
+            return detail::array_descriptor1_proxy(m_ptr)->names != nullptr;
+        }
+        const auto *proxy = detail::array_descriptor2_proxy(m_ptr);
+        if (proxy->type_num < 0 || proxy->type_num >= 2056) {
+            return false;
+        }
+        return proxy->names != nullptr;
+    }
+#endif
 
     /// Single-character code for dtype's kind.
     /// For example, floating point types are 'f' and integral types are 'i'.
@@ -596,14 +725,50 @@ class dtype : public object {
         return detail::array_descriptor_proxy(m_ptr)->type;
     }
 
+    /// type number of dtype.
+    int num() const {
+        // Note: The signature, `dtype::num` follows the naming of NumPy's public
+        // Python API (i.e., ``dtype.num``), rather than its internal
+        // C API (``PyArray_Descr::type_num``).
+        return detail::array_descriptor_proxy(m_ptr)->type_num;
+    }
+
+    /// Single character for byteorder
+    char byteorder() const { return detail::array_descriptor_proxy(m_ptr)->byteorder; }
+
+/// Alignment of the data type
+#ifdef PYBIND11_NUMPY_1_ONLY
+    int alignment() const { return detail::array_descriptor_proxy(m_ptr)->alignment; }
+#else
+    ssize_t alignment() const {
+        if (detail::npy_api::get().PyArray_RUNTIME_VERSION_ < 0x12) {
+            return detail::array_descriptor1_proxy(m_ptr)->alignment;
+        }
+        return detail::array_descriptor2_proxy(m_ptr)->alignment;
+    }
+#endif
+
+/// Flags for the array descriptor
+#ifdef PYBIND11_NUMPY_1_ONLY
+    char flags() const { return detail::array_descriptor_proxy(m_ptr)->flags; }
+#else
+    std::uint64_t flags() const {
+        if (detail::npy_api::get().PyArray_RUNTIME_VERSION_ < 0x12) {
+            return (unsigned char) detail::array_descriptor1_proxy(m_ptr)->flags;
+        }
+        return detail::array_descriptor2_proxy(m_ptr)->flags;
+    }
+#endif
+
 private:
-    static object _dtype_from_pep3118() {
-        static PyObject *obj = module_::import("numpy.core._internal")
-                                   .attr("_dtype_from_pep3118")
-                                   .cast<object>()
-                                   .release()
-                                   .ptr();
-        return reinterpret_borrow<object>(obj);
+    static object &_dtype_from_pep3118() {
+        PYBIND11_CONSTINIT static gil_safe_call_once_and_store<object> storage;
+        return storage
+            .call_once_and_store_result([]() {
+                return detail::import_numpy_core_submodule("_internal")
+                    .attr("_dtype_from_pep3118");
+            })
+            .get_stored();
     }
 
     dtype strip_padding(ssize_t itemsize) {
@@ -614,22 +779,27 @@ class dtype : public object {
         }
 
         struct field_descr {
-            PYBIND11_STR_TYPE name;
+            pybind11::str name;
             object format;
             pybind11::int_ offset;
+            field_descr(pybind11::str &&name, object &&format, pybind11::int_ &&offset)
+                : name{std::move(name)}, format{std::move(format)}, offset{std::move(offset)} {};
         };
+        auto field_dict = attr("fields").cast<dict>();
         std::vector<field_descr> field_descriptors;
+        field_descriptors.reserve(field_dict.size());
 
-        for (auto field : attr("fields").attr("items")()) {
+        for (auto field : field_dict.attr("items")()) {
             auto spec = field.cast<tuple>();
             auto name = spec[0].cast<pybind11::str>();
-            auto format = spec[1].cast<tuple>()[0].cast<dtype>();
-            auto offset = spec[1].cast<tuple>()[1].cast<pybind11::int_>();
+            auto spec_fo = spec[1].cast<tuple>();
+            auto format = spec_fo[0].cast<dtype>();
+            auto offset = spec_fo[1].cast<pybind11::int_>();
             if ((len(name) == 0u) && format.kind() == 'V') {
                 continue;
             }
-            field_descriptors.push_back(
-                {(PYBIND11_STR_TYPE) name, format.strip_padding(format.itemsize()), offset});
+            field_descriptors.emplace_back(
+                std::move(name), format.strip_padding(format.itemsize()), std::move(offset));
         }
 
         std::sort(field_descriptors.begin(),
@@ -640,9 +810,9 @@ class dtype : public object {
 
         list names, formats, offsets;
         for (auto &descr : field_descriptors) {
-            names.append(descr.name);
-            formats.append(descr.format);
-            offsets.append(descr.offset);
+            names.append(std::move(descr.name));
+            formats.append(std::move(descr.format));
+            offsets.append(std::move(descr.offset));
         }
         return dtype(std::move(names), std::move(formats), std::move(offsets), itemsize);
     }
@@ -755,9 +925,7 @@ class array : public buffer {
     }
 
     /// Byte size of a single element
-    ssize_t itemsize() const {
-        return detail::array_descriptor_proxy(detail::array_proxy(m_ptr)->descr)->elsize;
-    }
+    ssize_t itemsize() const { return dtype().itemsize(); }
 
     /// Total number of bytes
     ssize_t nbytes() const { return size() * itemsize(); }
@@ -846,7 +1014,7 @@ class array : public buffer {
      */
     template <typename T, ssize_t Dims = -1>
     detail::unchecked_mutable_reference<T, Dims> mutable_unchecked() & {
-        if (PYBIND11_SILENCE_MSVC_C4127(Dims >= 0) && ndim() != Dims) {
+        if (Dims >= 0 && ndim() != Dims) {
             throw std::domain_error("array has incorrect number of dimensions: "
                                     + std::to_string(ndim()) + "; expected "
                                     + std::to_string(Dims));
@@ -864,7 +1032,7 @@ class array : public buffer {
      */
     template <typename T, ssize_t Dims = -1>
     detail::unchecked_reference<T, Dims> unchecked() const & {
-        if (PYBIND11_SILENCE_MSVC_C4127(Dims >= 0) && ndim() != Dims) {
+        if (Dims >= 0 && ndim() != Dims) {
             throw std::domain_error("array has incorrect number of dimensions: "
                                     + std::to_string(ndim()) + "; expected "
                                     + std::to_string(Dims));
@@ -975,7 +1143,7 @@ class array : public buffer {
     /// Create array from any object -- always returns a new reference
     static PyObject *raw_array(PyObject *ptr, int ExtraFlags = 0) {
         if (ptr == nullptr) {
-            PyErr_SetString(PyExc_ValueError, "cannot create a pybind11::array from a nullptr");
+            set_error(PyExc_ValueError, "cannot create a pybind11::array from a nullptr");
             return nullptr;
         }
         return detail::npy_api::get().PyArray_FromAny_(
@@ -1090,10 +1258,10 @@ class array_t : public array {
 
     /**
      * Returns a proxy object that provides const access to the array's data without bounds or
-     * dimensionality checking.  Unlike `unchecked()`, this does not require that the underlying
-     * array have the `writable` flag.  Use with care: the array must not be destroyed or reshaped
-     * for the duration of the returned object, and the caller must take care not to access invalid
-     * dimensions or dimension indices.
+     * dimensionality checking.  Unlike `mutable_unchecked()`, this does not require that the
+     * underlying array have the `writable` flag.  Use with care: the array must not be destroyed
+     * or reshaped for the duration of the returned object, and the caller must take care not to
+     * access invalid dimensions or dimension indices.
      */
     template <ssize_t Dims = -1>
     detail::unchecked_reference<T, Dims> unchecked() const & {
@@ -1122,7 +1290,7 @@ class array_t : public array {
     /// Create array from any object -- always returns a new reference
     static PyObject *raw_array_t(PyObject *ptr) {
         if (ptr == nullptr) {
-            PyErr_SetString(PyExc_ValueError, "cannot create a pybind11::array_t from a nullptr");
+            set_error(PyExc_ValueError, "cannot create a pybind11::array_t from a nullptr");
             return nullptr;
         }
         return detail::npy_api::get().PyArray_FromAny_(ptr,
@@ -1252,12 +1420,16 @@ struct npy_format_descriptor<
 public:
     static constexpr int value = values[detail::is_fmt_numeric<T>::index];
 
-    static pybind11::dtype dtype() {
-        if (auto *ptr = npy_api::get().PyArray_DescrFromType_(value)) {
-            return reinterpret_steal<pybind11::dtype>(ptr);
-        }
-        pybind11_fail("Unsupported buffer format!");
-    }
+    static pybind11::dtype dtype() { return pybind11::dtype(/*typenum*/ value); }
+};
+
+template <typename T>
+struct npy_format_descriptor<T, enable_if_t<is_same_ignoring_cvref<T, PyObject *>::value>> {
+    static constexpr auto name = const_name("object");
+
+    static constexpr int value = npy_api::NPY_OBJECT_;
+
+    static pybind11::dtype dtype() { return pybind11::dtype(/*typenum*/ value); }
 };
 
 #define PYBIND11_DECL_CHAR_FMT                                                                    \
@@ -1335,7 +1507,7 @@ PYBIND11_NOINLINE void register_structured_dtype(any_container<field_descriptor>
             pybind11_fail(std::string("NumPy: unsupported field dtype: `") + field.name + "` @ "
                           + tinfo.name());
         }
-        names.append(PYBIND11_STR_TYPE(field.name));
+        names.append(pybind11::str(field.name));
         formats.append(field.descr);
         offsets.append(pybind11::int_(field.offset));
     }
@@ -1372,7 +1544,7 @@ PYBIND11_NOINLINE void register_structured_dtype(any_container<field_descriptor>
     oss << '}';
     auto format_str = oss.str();
 
-    // Sanity check: verify that NumPy properly parses our buffer format string
+    // Smoke test: verify that NumPy properly parses our buffer format string
     auto &api = npy_api::get();
     auto arr = array(buffer_info(nullptr, itemsize, format_str, 1));
     if (!api.PyArray_EquivTypes_(dtype_ptr, arr.dtype().ptr())) {
@@ -1380,8 +1552,10 @@ PYBIND11_NOINLINE void register_structured_dtype(any_container<field_descriptor>
     }
 
     auto tindex = std::type_index(tinfo);
-    numpy_internals.registered_dtypes[tindex] = {dtype_ptr, format_str};
-    get_internals().direct_conversions[tindex].push_back(direct_converter);
+    numpy_internals.registered_dtypes[tindex] = {dtype_ptr, std::move(format_str)};
+    with_internals([tindex, &direct_converter](internals &internals) {
+        internals.direct_conversions[tindex].push_back(direct_converter);
+    });
 }
 
 template <typename T, typename SFINAE>
@@ -1440,7 +1614,7 @@ struct npy_format_descriptor {
         }
 
 // Extract name, offset and format descriptor for a struct field
-#    define PYBIND11_FIELD_DESCRIPTOR(T, Field) PYBIND11_FIELD_DESCRIPTOR_EX(T, Field, #    Field)
+#    define PYBIND11_FIELD_DESCRIPTOR(T, Field) PYBIND11_FIELD_DESCRIPTOR_EX(T, Field, #Field)
 
 // The main idea of this macro is borrowed from https://github.com/swansontec/map-macro
 // (C) William Swanson, Paul Fultz
@@ -1527,7 +1701,7 @@ class common_iterator {
     void *data() const { return p_ptr; }
 
 private:
-    char *p_ptr{0};
+    char *p_ptr{nullptr};
     container_type m_strides;
 };
 
@@ -1837,8 +2011,13 @@ struct vectorize_helper {
 
         auto result = returned_array::create(trivial, shape);
 
+        PYBIND11_WARNING_PUSH
+#ifdef PYBIND11_DETECTED_CLANG_WITH_MISLEADING_CALL_STD_MOVE_EXPLICITLY_WARNING
+        PYBIND11_WARNING_DISABLE_CLANG("-Wreturn-std-move")
+#endif
+
         if (size == 0) {
-            return std::move(result);
+            return result;
         }
 
         /* Call the function */
@@ -1849,7 +2028,8 @@ struct vectorize_helper {
             apply_trivial(buffers, params, mutable_data, size, i_seq, vi_seq, bi_seq);
         }
 
-        return std::move(result);
+        return result;
+        PYBIND11_WARNING_POP
     }
 
     template <size_t... Index, size_t... VIndex, size_t... BIndex>
diff --git a/include/pybind11/operators.h b/include/pybind11/operators.h
index a0c3b78d65..16a88ae171 100644
--- a/include/pybind11/operators.h
+++ b/include/pybind11/operators.h
@@ -84,6 +84,7 @@ struct op_impl {};
 /// Operator implementation generator
 template <op_id id, op_type ot, typename L, typename R>
 struct op_ {
+    static constexpr bool op_enable_if_hook = true;
     template <typename Class, typename... Extra>
     void execute(Class &cl, const Extra &...extra) const {
         using Base = typename Class::type;
diff --git a/include/pybind11/options.h b/include/pybind11/options.h
index 1e493bdcc4..1b2122522d 100644
--- a/include/pybind11/options.h
+++ b/include/pybind11/options.h
@@ -47,6 +47,16 @@ class options {
         return *this;
     }
 
+    options &disable_enum_members_docstring() & {
+        global_state().show_enum_members_docstring = false;
+        return *this;
+    }
+
+    options &enable_enum_members_docstring() & {
+        global_state().show_enum_members_docstring = true;
+        return *this;
+    }
+
     // Getter methods (return the global state):
 
     static bool show_user_defined_docstrings() {
@@ -55,6 +65,10 @@ class options {
 
     static bool show_function_signatures() { return global_state().show_function_signatures; }
 
+    static bool show_enum_members_docstring() {
+        return global_state().show_enum_members_docstring;
+    }
+
     // This type is not meant to be allocated on the heap.
     void *operator new(size_t) = delete;
 
@@ -63,6 +77,8 @@ class options {
         bool show_user_defined_docstrings = true; //< Include user-supplied texts in docstrings.
         bool show_function_signatures = true;     //< Include auto-generated function signatures
                                                   //  in docstrings.
+        bool show_enum_members_docstring = true;  //< Include auto-generated member list in enum
+                                                  //  docstrings.
     };
 
     static state &global_state() {
diff --git a/include/pybind11/pybind11.h b/include/pybind11/pybind11.h
index fa018b509d..74919a7d52 100644
--- a/include/pybind11/pybind11.h
+++ b/include/pybind11/pybind11.h
@@ -14,7 +14,9 @@
 #include "detail/init.h"
 #include "attr.h"
 #include "gil.h"
+#include "gil_safe_call_once.h"
 #include "options.h"
+#include "typing.h"
 
 #include <cstdlib>
 #include <cstring>
@@ -35,6 +37,8 @@
 #    include <cxxabi.h>
 #endif
 
+PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)
+
 /* https://stackoverflow.com/questions/46798456/handling-gccs-noexcept-type-warning
    This warning is about ABI compatibility, not code health.
    It is only actually needed in a couple places, but apparently GCC 7 "generates this warning if
@@ -43,14 +47,54 @@
    No other GCC version generates this warning.
  */
 #if defined(__GNUC__) && __GNUC__ == 7
-#    pragma GCC diagnostic push
-#    pragma GCC diagnostic ignored "-Wnoexcept-type"
+PYBIND11_WARNING_DISABLE_GCC("-Wnoexcept-type")
 #endif
 
-PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)
+PYBIND11_WARNING_DISABLE_MSVC(4127)
 
 PYBIND11_NAMESPACE_BEGIN(detail)
 
+inline std::string replace_newlines_and_squash(const char *text) {
+    const char *whitespaces = " \t\n\r\f\v";
+    std::string result(text);
+    bool previous_is_whitespace = false;
+
+    if (result.size() >= 2) {
+        // Do not modify string representations
+        char first_char = result[0];
+        char last_char = result[result.size() - 1];
+        if (first_char == last_char && first_char == '\'') {
+            return result;
+        }
+    }
+    result.clear();
+
+    // Replace characters in whitespaces array with spaces and squash consecutive spaces
+    while (*text != '\0') {
+        if (std::strchr(whitespaces, *text)) {
+            if (!previous_is_whitespace) {
+                result += ' ';
+                previous_is_whitespace = true;
+            }
+        } else {
+            result += *text;
+            previous_is_whitespace = false;
+        }
+        ++text;
+    }
+
+    // Strip leading and trailing whitespaces
+    const size_t str_begin = result.find_first_not_of(whitespaces);
+    if (str_begin == std::string::npos) {
+        return "";
+    }
+
+    const size_t str_end = result.find_last_not_of(whitespaces);
+    const size_t str_range = str_end - str_begin + 1;
+
+    return result.substr(str_begin, str_range);
+}
+
 // Apply all the extensions translators from a list
 // Return true if one of the translators completed without raising an exception
 // itself. Return of false indicates that if there are other translators
@@ -83,6 +127,7 @@ class cpp_function : public function {
     cpp_function() = default;
     // NOLINTNEXTLINE(google-explicit-constructor)
     cpp_function(std::nullptr_t) {}
+    cpp_function(std::nullptr_t, const is_setter &) {}
 
     /// Construct a cpp_function from a vanilla function pointer
     template <typename Return, typename... Args, typename... Extra>
@@ -177,22 +222,22 @@ class cpp_function : public function {
         auto *rec = unique_rec.get();
 
         /* Store the capture object directly in the function record if there is enough space */
-        if (PYBIND11_SILENCE_MSVC_C4127(sizeof(capture) <= sizeof(rec->data))) {
+        if (sizeof(capture) <= sizeof(rec->data)) {
             /* Without these pragmas, GCC warns that there might not be
                enough space to use the placement new operator. However, the
                'if' statement above ensures that this is the case. */
-#if defined(__GNUG__) && __GNUC__ >= 6 && !defined(__clang__) && !defined(__INTEL_COMPILER)
-#    pragma GCC diagnostic push
-#    pragma GCC diagnostic ignored "-Wplacement-new"
+            PYBIND11_WARNING_PUSH
+
+#if defined(__GNUG__) && __GNUC__ >= 6
+            PYBIND11_WARNING_DISABLE_GCC("-Wplacement-new")
 #endif
+
             new ((capture *) &rec->data) capture{std::forward<Func>(f)};
-#if defined(__GNUG__) && __GNUC__ >= 6 && !defined(__clang__) && !defined(__INTEL_COMPILER)
-#    pragma GCC diagnostic pop
-#endif
-#if defined(__GNUG__) && !PYBIND11_HAS_STD_LAUNDER && !defined(__INTEL_COMPILER)
-#    pragma GCC diagnostic push
-#    pragma GCC diagnostic ignored "-Wstrict-aliasing"
+
+#if !PYBIND11_HAS_STD_LAUNDER
+            PYBIND11_WARNING_DISABLE_GCC("-Wstrict-aliasing")
 #endif
+
             // UB without std::launder, but without breaking ABI and/or
             // a significant refactoring it's "impossible" to solve.
             if (!std::is_trivially_destructible<capture>::value) {
@@ -202,9 +247,7 @@ class cpp_function : public function {
                     data->~capture();
                 };
             }
-#if defined(__GNUG__) && !PYBIND11_HAS_STD_LAUNDER && !defined(__INTEL_COMPILER)
-#    pragma GCC diagnostic pop
-#endif
+            PYBIND11_WARNING_POP
         } else {
             rec->data[0] = new capture{std::forward<Func>(f)};
             rec->free_data = [](function_record *r) { delete ((capture *) r->data[0]); };
@@ -245,10 +288,16 @@ class cpp_function : public function {
             using Guard = extract_guard_t<Extra...>;
 
             /* Perform the function call */
-            handle result
-                = cast_out::cast(std::move(args_converter).template call<Return, Guard>(cap->f),
-                                 policy,
-                                 call.parent);
+            handle result;
+            if (call.func.is_setter) {
+                (void) std::move(args_converter).template call<Return, Guard>(cap->f);
+                result = none().release();
+            } else {
+                result = cast_out::cast(
+                    std::move(args_converter).template call<Return, Guard>(cap->f),
+                    policy,
+                    call.parent);
+            }
 
             /* Invoke call policy post-call hook */
             process_attributes<Extra...>::postcall(call, result);
@@ -312,6 +361,10 @@ class cpp_function : public function {
     // along the way.
     class strdup_guard {
     public:
+        strdup_guard() = default;
+        strdup_guard(const strdup_guard &) = delete;
+        strdup_guard &operator=(const strdup_guard &) = delete;
+
         ~strdup_guard() {
             for (auto *s : strings) {
                 std::free(s);
@@ -366,7 +419,7 @@ class cpp_function : public function {
         rec->is_constructor = (std::strcmp(rec->name, "__init__") == 0)
                               || (std::strcmp(rec->name, "__setstate__") == 0);
 
-#if !defined(NDEBUG) && !defined(PYBIND11_DISABLE_NEW_STYLE_INIT_WARNING)
+#if defined(PYBIND11_DETAILED_ERROR_MESSAGES) && !defined(PYBIND11_DISABLE_NEW_STYLE_INIT_WARNING)
         if (rec->is_constructor && !rec->is_new_style_constructor) {
             const auto class_name
                 = detail::get_fully_qualified_tp_name((PyTypeObject *) rec->scope.ptr());
@@ -414,7 +467,7 @@ class cpp_function : public function {
                 // Write default value if available.
                 if (!is_starred && arg_index < rec->args.size() && rec->args[arg_index].descr) {
                     signature += " = ";
-                    signature += rec->args[arg_index].descr;
+                    signature += detail::replace_newlines_and_squash(rec->args[arg_index].descr);
                 }
                 // Separator for positional-only arguments (placed after the
                 // argument, rather than before like *
@@ -439,9 +492,7 @@ class cpp_function : public function {
                     signature += rec->scope.attr("__module__").cast<std::string>() + "."
                                  + rec->scope.attr("__qualname__").cast<std::string>();
                 } else {
-                    std::string tname(t->name());
-                    detail::clean_type_id(tname);
-                    signature += tname;
+                    signature += detail::quote_cpp_type_name(detail::clean_type_id(t->name()));
                 }
             } else {
                 signature += c;
@@ -464,13 +515,20 @@ class cpp_function : public function {
         if (rec->sibling) {
             if (PyCFunction_Check(rec->sibling.ptr())) {
                 auto *self = PyCFunction_GET_SELF(rec->sibling.ptr());
-                capsule rec_capsule = isinstance<capsule>(self) ? reinterpret_borrow<capsule>(self)
-                                                                : capsule(self);
-                chain = (detail::function_record *) rec_capsule;
-                /* Never append a method to an overload chain of a parent class;
-                   instead, hide the parent's overloads in this case */
-                if (!chain->scope.is(rec->scope)) {
+                if (!isinstance<capsule>(self)) {
                     chain = nullptr;
+                } else {
+                    auto rec_capsule = reinterpret_borrow<capsule>(self);
+                    if (detail::is_function_record_capsule(rec_capsule)) {
+                        chain = rec_capsule.get_pointer<detail::function_record>();
+                        /* Never append a method to an overload chain of a parent class;
+                           instead, hide the parent's overloads in this case */
+                        if (!chain->scope.is(rec->scope)) {
+                            chain = nullptr;
+                        }
+                    } else {
+                        chain = nullptr;
+                    }
                 }
             }
             // Don't trigger for things like the default __init__, which are wrapper_descriptors
@@ -491,6 +549,7 @@ class cpp_function : public function {
             rec->def->ml_flags = METH_VARARGS | METH_KEYWORDS;
 
             capsule rec_capsule(unique_rec.release(),
+                                detail::get_function_record_capsule_name(),
                                 [](void *ptr) { destruct((detail::function_record *) ptr); });
             guarded_strdup.release();
 
@@ -514,8 +573,9 @@ class cpp_function : public function {
             if (chain->is_method != rec->is_method) {
                 pybind11_fail(
                     "overloading a method with both static and instance methods is not supported; "
-#if defined(NDEBUG)
-                    "compile in debug mode for more details"
+#if !defined(PYBIND11_DETAILED_ERROR_MESSAGES)
+                    "#define PYBIND11_DETAILED_ERROR_MESSAGES or compile in debug mode for more "
+                    "details"
 #else
                     "error while attempting to bind "
                     + std::string(rec->is_method ? "instance" : "static") + " method "
@@ -656,10 +716,13 @@ class cpp_function : public function {
     /// Main dispatch logic for calls to functions bound using pybind11
     static PyObject *dispatcher(PyObject *self, PyObject *args_in, PyObject *kwargs_in) {
         using namespace detail;
+        assert(isinstance<capsule>(self));
 
         /* Iterator over the list of potentially admissible overloads */
-        const function_record *overloads = (function_record *) PyCapsule_GetPointer(self, nullptr),
-                              *it = overloads;
+        const function_record *overloads = reinterpret_cast<function_record *>(
+                                  PyCapsule_GetPointer(self, get_function_record_capsule_name())),
+                              *current_overload = overloads;
+        assert(overloads != nullptr);
 
         /* Need to know how many arguments + keyword arguments there are to pick the right
            overload */
@@ -672,9 +735,8 @@ class cpp_function : public function {
         if (overloads->is_constructor) {
             if (!parent
                 || !PyObject_TypeCheck(parent.ptr(), (PyTypeObject *) overloads->scope.ptr())) {
-                PyErr_SetString(
-                    PyExc_TypeError,
-                    "__init__(self, ...) called with invalid or missing `self` argument");
+                set_error(PyExc_TypeError,
+                          "__init__(self, ...) called with invalid or missing `self` argument");
                 return nullptr;
             }
 
@@ -697,9 +759,10 @@ class cpp_function : public function {
             std::vector<function_call> second_pass;
 
             // However, if there are no overloads, we can just skip the no-convert pass entirely
-            const bool overloaded = it != nullptr && it->next != nullptr;
+            const bool overloaded
+                = current_overload != nullptr && current_overload->next != nullptr;
 
-            for (; it != nullptr; it = it->next) {
+            for (; current_overload != nullptr; current_overload = current_overload->next) {
 
                 /* For each overload:
                    1. Copy all positional arguments we were given, also checking to make sure that
@@ -720,7 +783,7 @@ class cpp_function : public function {
                    a result other than PYBIND11_TRY_NEXT_OVERLOAD.
                  */
 
-                const function_record &func = *it;
+                const function_record &func = *current_overload;
                 size_t num_args = func.nargs; // Number of positional arguments that we need
                 if (func.has_args) {
                     --num_args; // (but don't count py::args
@@ -902,7 +965,7 @@ class cpp_function : public function {
 
 // 5. Put everything in a vector.  Not technically step 5, we've been building it
 // in `call.args` all along.
-#if !defined(NDEBUG)
+#if defined(PYBIND11_DETAILED_ERROR_MESSAGES)
                 if (call.args.size() != func.nargs || call.args_convert.size() != func.nargs) {
                     pybind11_fail("Internal error: function call dispatcher inserted wrong number "
                                   "of arguments!");
@@ -958,10 +1021,10 @@ class cpp_function : public function {
                     }
 
                     if (result.ptr() != PYBIND11_TRY_NEXT_OVERLOAD) {
-                        // The error reporting logic below expects 'it' to be valid, as it would be
-                        // if we'd encountered this failure in the first-pass loop.
+                        // The error reporting logic below expects 'current_overload' to be valid,
+                        // as it would be if we'd encountered this failure in the first-pass loop.
                         if (!result) {
-                            it = &call.func;
+                            current_overload = &call.func;
                         }
                         break;
                     }
@@ -985,24 +1048,30 @@ class cpp_function : public function {
 
                A translator may choose to do one of the following:
 
-                - catch the exception and call PyErr_SetString or PyErr_SetObject
+                - catch the exception and call py::set_error()
                   to set a standard (or custom) Python exception, or
                 - do nothing and let the exception fall through to the next translator, or
                 - delegate translation to the next translator by throwing a new type of exception.
              */
 
-            auto &local_exception_translators
-                = get_local_internals().registered_exception_translators;
-            if (detail::apply_exception_translators(local_exception_translators)) {
-                return nullptr;
-            }
-            auto &exception_translators = get_internals().registered_exception_translators;
-            if (detail::apply_exception_translators(exception_translators)) {
+            bool handled = with_internals([&](internals &internals) {
+                auto &local_exception_translators
+                    = get_local_internals().registered_exception_translators;
+                if (detail::apply_exception_translators(local_exception_translators)) {
+                    return true;
+                }
+                auto &exception_translators = internals.registered_exception_translators;
+                if (detail::apply_exception_translators(exception_translators)) {
+                    return true;
+                }
+                return false;
+            });
+
+            if (handled) {
                 return nullptr;
             }
 
-            PyErr_SetString(PyExc_SystemError,
-                            "Exception escaped from default exception translator!");
+            set_error(PyExc_SystemError, "Exception escaped from default exception translator!");
             return nullptr;
         }
 
@@ -1080,7 +1149,7 @@ class cpp_function : public function {
                     }
                     msg += "kwargs: ";
                     bool first = true;
-                    for (auto kwarg : kwargs) {
+                    for (const auto &kwarg : kwargs) {
                         if (first) {
                             first = false;
                         } else {
@@ -1103,20 +1172,21 @@ class cpp_function : public function {
                 raise_from(PyExc_TypeError, msg.c_str());
                 return nullptr;
             }
-            PyErr_SetString(PyExc_TypeError, msg.c_str());
+            set_error(PyExc_TypeError, msg.c_str());
             return nullptr;
         }
         if (!result) {
             std::string msg = "Unable to convert function return value to a "
                               "Python type! The signature was\n\t";
-            msg += it->signature;
+            assert(current_overload != nullptr);
+            msg += current_overload->signature;
             append_note_if_missing_header_is_suspected(msg);
             // Attach additional error info to the exception if supported
             if (PyErr_Occurred()) {
                 raise_from(PyExc_TypeError, msg.c_str());
                 return nullptr;
             }
-            PyErr_SetString(PyExc_TypeError, msg.c_str());
+            set_error(PyExc_TypeError, msg.c_str());
             return nullptr;
         }
         if (overloads->is_constructor && !self_value_and_holder.holder_constructed()) {
@@ -1127,6 +1197,25 @@ class cpp_function : public function {
     }
 };
 
+PYBIND11_NAMESPACE_BEGIN(detail)
+
+template <>
+struct handle_type_name<cpp_function> {
+    static constexpr auto name = const_name("Callable");
+};
+
+PYBIND11_NAMESPACE_END(detail)
+
+// Use to activate Py_MOD_GIL_NOT_USED.
+class mod_gil_not_used {
+public:
+    explicit mod_gil_not_used(bool flag = true) : flag_(flag) {}
+    bool flag() const { return flag_; }
+
+private:
+    bool flag_;
+};
+
 /// Wrapper for Python extension modules
 class module_ : public object {
 public:
@@ -1168,9 +1257,16 @@ class module_ : public object {
             py::module_ m3 = m2.def_submodule("subsub", "A submodule of 'example.sub'");
     \endrst */
     module_ def_submodule(const char *name, const char *doc = nullptr) {
-        std::string full_name
-            = std::string(PyModule_GetName(m_ptr)) + std::string(".") + std::string(name);
-        auto result = reinterpret_borrow<module_>(PyImport_AddModule(full_name.c_str()));
+        const char *this_name = PyModule_GetName(m_ptr);
+        if (this_name == nullptr) {
+            throw error_already_set();
+        }
+        std::string full_name = std::string(this_name) + '.' + name;
+        handle submodule = PyImport_AddModule(full_name.c_str());
+        if (!submodule) {
+            throw error_already_set();
+        }
+        auto result = reinterpret_borrow<module_>(submodule);
         if (doc && options::show_user_defined_docstrings()) {
             result.attr("__doc__") = pybind11::str(doc);
         }
@@ -1220,7 +1316,11 @@ class module_ : public object {
 
         ``def`` should point to a statically allocated module_def.
     \endrst */
-    static module_ create_extension_module(const char *name, const char *doc, module_def *def) {
+    static module_ create_extension_module(const char *name,
+                                           const char *doc,
+                                           module_def *def,
+                                           mod_gil_not_used gil_not_used
+                                           = mod_gil_not_used(false)) {
         // module_def is PyModuleDef
         // Placement new (not an allocation).
         def = new (def)
@@ -1240,6 +1340,11 @@ class module_ : public object {
             }
             pybind11_fail("Internal error in module_::create_extension_module()");
         }
+        if (gil_not_used.flag()) {
+#ifdef Py_GIL_DISABLED
+            PyUnstable_Module_SetGIL(m, Py_MOD_GIL_NOT_USED);
+#endif
+        }
         // TODO: Should be reinterpret_steal for Python 3, but Python also steals it again when
         //       returned from PyInit_...
         //       For Python 2, reinterpret_borrow was correct.
@@ -1247,6 +1352,15 @@ class module_ : public object {
     }
 };
 
+PYBIND11_NAMESPACE_BEGIN(detail)
+
+template <>
+struct handle_type_name<module_> {
+    static constexpr auto name = const_name("module");
+};
+
+PYBIND11_NAMESPACE_END(detail)
+
 // When inside a namespace (or anywhere as long as it's not the first item on a line),
 // C++20 allows "module" to be used. This is provided for backward compatibility, and for
 // simplicity, if someone wants to use py::module for example, that is perfectly safe.
@@ -1256,8 +1370,14 @@ using module = module_;
 /// Return a dictionary representing the global variables in the current execution frame,
 /// or ``__main__.__dict__`` if there is no frame (usually when the interpreter is embedded).
 inline dict globals() {
+#if PY_VERSION_HEX >= 0x030d0000
+    PyObject *p = PyEval_GetFrameGlobals();
+    return p ? reinterpret_steal<dict>(p)
+             : reinterpret_borrow<dict>(module_::import("__main__").attr("__dict__").ptr());
+#else
     PyObject *p = PyEval_GetGlobals();
     return reinterpret_borrow<dict>(p ? p : module_::import("__main__").attr("__dict__").ptr());
+#endif
 }
 
 template <typename... Args, typename = detail::enable_if_t<args_are_all_keyword_or_ds<Args...>()>>
@@ -1303,15 +1423,16 @@ class generic_type : public object {
         tinfo->default_holder = rec.default_holder;
         tinfo->module_local = rec.module_local;
 
-        auto &internals = get_internals();
-        auto tindex = std::type_index(*rec.type);
-        tinfo->direct_conversions = &internals.direct_conversions[tindex];
-        if (rec.module_local) {
-            get_local_internals().registered_types_cpp[tindex] = tinfo;
-        } else {
-            internals.registered_types_cpp[tindex] = tinfo;
-        }
-        internals.registered_types_py[(PyTypeObject *) m_ptr] = {tinfo};
+        with_internals([&](internals &internals) {
+            auto tindex = std::type_index(*rec.type);
+            tinfo->direct_conversions = &internals.direct_conversions[tindex];
+            if (rec.module_local) {
+                get_local_internals().registered_types_cpp[tindex] = tinfo;
+            } else {
+                internals.registered_types_cpp[tindex] = tinfo;
+            }
+            internals.registered_types_py[(PyTypeObject *) m_ptr] = {tinfo};
+        });
 
         if (rec.bases.size() > 1 || rec.multiple_inheritance) {
             mark_parents_nonsimple(tinfo->type);
@@ -1404,9 +1525,9 @@ template <typename T, enable_if_t<has_operator_delete<T>::value, int> = 0>
 void call_operator_delete(T *p, size_t, size_t) {
     T::operator delete(p);
 }
-template <
-    typename T,
-    enable_if_t<!has_operator_delete<T>::value && has_operator_delete_size<T>::value, int> = 0>
+template <typename T,
+          enable_if_t<!has_operator_delete<T>::value && has_operator_delete_size<T>::value, int>
+          = 0>
 void call_operator_delete(T *p, size_t s, size_t) {
     T::operator delete(p, s);
 }
@@ -1524,10 +1645,12 @@ class class_ : public detail::generic_type {
         generic_type::initialize(record);
 
         if (has_alias) {
-            auto &instances = record.module_local ? get_local_internals().registered_types_cpp
-                                                  : get_internals().registered_types_cpp;
-            instances[std::type_index(typeid(type_alias))]
-                = instances[std::type_index(typeid(type))];
+            with_internals([&](internals &internals) {
+                auto &instances = record.module_local ? get_local_internals().registered_types_cpp
+                                                      : internals.registered_types_cpp;
+                instances[std::type_index(typeid(type_alias))]
+                    = instances[std::type_index(typeid(type))];
+            });
         }
     }
 
@@ -1561,18 +1684,19 @@ class class_ : public detail::generic_type {
                         scope(*this),
                         sibling(getattr(*this, name_, none())),
                         extra...);
-        attr(cf.name()) = staticmethod(cf);
+        auto cf_name = cf.name();
+        attr(std::move(cf_name)) = staticmethod(std::move(cf));
         return *this;
     }
 
-    template <detail::op_id id, detail::op_type ot, typename L, typename R, typename... Extra>
-    class_ &def(const detail::op_<id, ot, L, R> &op, const Extra &...extra) {
+    template <typename T, typename... Extra, detail::enable_if_t<T::op_enable_if_hook, int> = 0>
+    class_ &def(const T &op, const Extra &...extra) {
         op.execute(*this, extra...);
         return *this;
     }
 
-    template <detail::op_id id, detail::op_type ot, typename L, typename R, typename... Extra>
-    class_ &def_cast(const detail::op_<id, ot, L, R> &op, const Extra &...extra) {
+    template <typename T, typename... Extra, detail::enable_if_t<T::op_enable_if_hook, int> = 0>
+    class_ &def_cast(const T &op, const Extra &...extra) {
         op.execute_cast(*this, extra...);
         return *this;
     }
@@ -1615,7 +1739,7 @@ class class_ : public detail::generic_type {
                 if (!caster.load(obj, false)) {
                     return nullptr;
                 }
-                return new buffer_info(((capture *) ptr)->func(caster));
+                return new buffer_info(((capture *) ptr)->func(std::move(caster)));
             },
             ptr);
         weakref(m_ptr, cpp_function([ptr](handle wr) {
@@ -1706,7 +1830,8 @@ class class_ : public detail::generic_type {
     template <typename Getter, typename Setter, typename... Extra>
     class_ &
     def_property(const char *name, const Getter &fget, const Setter &fset, const Extra &...extra) {
-        return def_property(name, fget, cpp_function(method_adaptor<type>(fset)), extra...);
+        return def_property(
+            name, fget, cpp_function(method_adaptor<type>(fset), is_setter()), extra...);
     }
     template <typename Getter, typename... Extra>
     class_ &def_property(const char *name,
@@ -1817,7 +1942,7 @@ class class_ : public detail::generic_type {
         if (holder_ptr) {
             init_holder_from_existing(v_h, holder_ptr, std::is_copy_constructible<holder_type>());
             v_h.set_holder_constructed();
-        } else if (inst->owned || detail::always_construct_holder<holder_type>::value) {
+        } else if (detail::always_construct_holder<holder_type>::value || inst->owned) {
             new (std::addressof(v_h.holder<holder_type>())) holder_type(v_h.value_ptr<type>());
             v_h.set_holder_constructed();
         }
@@ -1857,9 +1982,22 @@ class class_ : public detail::generic_type {
 
     static detail::function_record *get_function_record(handle h) {
         h = detail::get_function(h);
-        return h ? (detail::function_record *) reinterpret_borrow<capsule>(
-                   PyCFunction_GET_SELF(h.ptr()))
-                 : nullptr;
+        if (!h) {
+            return nullptr;
+        }
+
+        handle func_self = PyCFunction_GET_SELF(h.ptr());
+        if (!func_self) {
+            throw error_already_set();
+        }
+        if (!isinstance<capsule>(func_self)) {
+            return nullptr;
+        }
+        auto cap = reinterpret_borrow<capsule>(func_self);
+        if (!detail::is_function_record_capsule(cap)) {
+            return nullptr;
+        }
+        return cap.get_pointer<detail::function_record>();
     }
 };
 
@@ -1920,7 +2058,8 @@ struct enum_base {
             [](const object &arg) -> str {
                 handle type = type::handle_of(arg);
                 object type_name = type.attr("__name__");
-                return pybind11::str("<{}.{}: {}>").format(type_name, enum_name(arg), int_(arg));
+                return pybind11::str("<{}.{}: {}>")
+                    .format(std::move(type_name), enum_name(arg), int_(arg));
             },
             name("__repr__"),
             is_method(m_base));
@@ -1930,34 +2069,40 @@ struct enum_base {
         m_base.attr("__str__") = cpp_function(
             [](handle arg) -> str {
                 object type_name = type::handle_of(arg).attr("__name__");
-                return pybind11::str("{}.{}").format(type_name, enum_name(arg));
+                return pybind11::str("{}.{}").format(std::move(type_name), enum_name(arg));
             },
-            name("name"),
+            name("__str__"),
             is_method(m_base));
 
-        m_base.attr("__doc__") = static_property(
-            cpp_function(
-                [](handle arg) -> std::string {
-                    std::string docstring;
-                    dict entries = arg.attr("__entries");
-                    if (((PyTypeObject *) arg.ptr())->tp_doc) {
-                        docstring += std::string(((PyTypeObject *) arg.ptr())->tp_doc) + "\n\n";
-                    }
-                    docstring += "Members:";
-                    for (auto kv : entries) {
-                        auto key = std::string(pybind11::str(kv.first));
-                        auto comment = kv.second[int_(1)];
-                        docstring += "\n\n  " + key;
-                        if (!comment.is_none()) {
-                            docstring += " : " + (std::string) pybind11::str(comment);
+        if (options::show_enum_members_docstring()) {
+            m_base.attr("__doc__") = static_property(
+                cpp_function(
+                    [](handle arg) -> std::string {
+                        std::string docstring;
+                        dict entries = arg.attr("__entries");
+                        if (((PyTypeObject *) arg.ptr())->tp_doc) {
+                            docstring += std::string(
+                                reinterpret_cast<PyTypeObject *>(arg.ptr())->tp_doc);
+                            docstring += "\n\n";
                         }
-                    }
-                    return docstring;
-                },
-                name("__doc__")),
-            none(),
-            none(),
-            "");
+                        docstring += "Members:";
+                        for (auto kv : entries) {
+                            auto key = std::string(pybind11::str(kv.first));
+                            auto comment = kv.second[int_(1)];
+                            docstring += "\n\n  ";
+                            docstring += key;
+                            if (!comment.is_none()) {
+                                docstring += " : ";
+                                docstring += pybind11::str(comment).cast<std::string>();
+                            }
+                        }
+                        return docstring;
+                    },
+                    name("__doc__")),
+                none(),
+                none(),
+                "");
+        }
 
         m_base.attr("__members__") = static_property(cpp_function(
                                                          [](handle arg) -> dict {
@@ -2054,12 +2199,12 @@ struct enum_base {
         str name(name_);
         if (entries.contains(name)) {
             std::string type_name = (std::string) str(m_base.attr("__name__"));
-            throw value_error(type_name + ": element \"" + std::string(name_)
+            throw value_error(std::move(type_name) + ": element \"" + std::string(name_)
                               + "\" already exists!");
         }
 
-        entries[name] = std::make_pair(value, doc);
-        m_base.attr(name) = value;
+        entries[name] = pybind11::make_tuple(value, doc);
+        m_base.attr(std::move(name)) = std::move(value);
     }
 
     PYBIND11_NOINLINE void export_values() {
@@ -2220,28 +2365,32 @@ 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 = get_internals()
-                   .registered_types_py
+    auto res = with_internals([type](internals &internals) {
+        return internals
+            .registered_types_py
 #ifdef __cpp_lib_unordered_map_try_emplace
-                   .try_emplace(type);
+            .try_emplace(type);
 #else
-                   .emplace(type, std::vector<detail::type_info *>());
+            .emplace(type, std::vector<detail::type_info *>());
 #endif
+    });
     if (res.second) {
         // New cache entry created; set up a weak reference to automatically remove it if the type
         // gets destroyed:
         weakref((PyObject *) type, cpp_function([type](handle wr) {
-                    get_internals().registered_types_py.erase(type);
-
-                    // TODO consolidate the erasure code in pybind11_meta_dealloc() in class.h
-                    auto &cache = get_internals().inactive_override_cache;
-                    for (auto it = cache.begin(), last = cache.end(); it != last;) {
-                        if (it->first == reinterpret_cast<PyObject *>(type)) {
-                            it = cache.erase(it);
-                        } else {
-                            ++it;
+                    with_internals([type](internals &internals) {
+                        internals.registered_types_py.erase(type);
+
+                        // TODO consolidate the erasure code in pybind11_meta_dealloc() in class.h
+                        auto &cache = internals.inactive_override_cache;
+                        for (auto it = cache.begin(), last = cache.end(); it != last;) {
+                            if (it->first == reinterpret_cast<PyObject *>(type)) {
+                                it = cache.erase(it);
+                            } else {
+                                ++it;
+                            }
                         }
-                    }
+                    });
 
                     wr.dec_ref();
                 }))
@@ -2344,7 +2493,7 @@ iterator make_iterator_impl(Iterator first, Sentinel last, Extra &&...extra) {
                 Policy);
     }
 
-    return cast(state{first, last, true});
+    return cast(state{std::forward<Iterator>(first), std::forward<Sentinel>(last), true});
 }
 
 PYBIND11_NAMESPACE_END(detail)
@@ -2355,13 +2504,15 @@ template <return_value_policy Policy = return_value_policy::reference_internal,
           typename Sentinel,
           typename ValueType = typename detail::iterator_access<Iterator>::result_type,
           typename... Extra>
-iterator make_iterator(Iterator first, Sentinel last, Extra &&...extra) {
+typing::Iterator<ValueType> make_iterator(Iterator first, Sentinel last, Extra &&...extra) {
     return detail::make_iterator_impl<detail::iterator_access<Iterator>,
                                       Policy,
                                       Iterator,
                                       Sentinel,
                                       ValueType,
-                                      Extra...>(first, last, std::forward<Extra>(extra)...);
+                                      Extra...>(std::forward<Iterator>(first),
+                                                std::forward<Sentinel>(last),
+                                                std::forward<Extra>(extra)...);
 }
 
 /// Makes a python iterator over the keys (`.first`) of a iterator over pairs from a
@@ -2371,13 +2522,15 @@ template <return_value_policy Policy = return_value_policy::reference_internal,
           typename Sentinel,
           typename KeyType = typename detail::iterator_key_access<Iterator>::result_type,
           typename... Extra>
-iterator make_key_iterator(Iterator first, Sentinel last, Extra &&...extra) {
+typing::Iterator<KeyType> make_key_iterator(Iterator first, Sentinel last, Extra &&...extra) {
     return detail::make_iterator_impl<detail::iterator_key_access<Iterator>,
                                       Policy,
                                       Iterator,
                                       Sentinel,
                                       KeyType,
-                                      Extra...>(first, last, std::forward<Extra>(extra)...);
+                                      Extra...>(std::forward<Iterator>(first),
+                                                std::forward<Sentinel>(last),
+                                                std::forward<Extra>(extra)...);
 }
 
 /// Makes a python iterator over the values (`.second`) of a iterator over pairs from a
@@ -2387,40 +2540,51 @@ template <return_value_policy Policy = return_value_policy::reference_internal,
           typename Sentinel,
           typename ValueType = typename detail::iterator_value_access<Iterator>::result_type,
           typename... Extra>
-iterator make_value_iterator(Iterator first, Sentinel last, Extra &&...extra) {
+typing::Iterator<ValueType> make_value_iterator(Iterator first, Sentinel last, Extra &&...extra) {
     return detail::make_iterator_impl<detail::iterator_value_access<Iterator>,
                                       Policy,
                                       Iterator,
                                       Sentinel,
                                       ValueType,
-                                      Extra...>(first, last, std::forward<Extra>(extra)...);
+                                      Extra...>(std::forward<Iterator>(first),
+                                                std::forward<Sentinel>(last),
+                                                std::forward<Extra>(extra)...);
 }
 
 /// Makes an iterator over values of an stl container or other container supporting
 /// `std::begin()`/`std::end()`
 template <return_value_policy Policy = return_value_policy::reference_internal,
           typename Type,
+          typename ValueType = typename detail::iterator_access<
+              decltype(std::begin(std::declval<Type &>()))>::result_type,
           typename... Extra>
-iterator make_iterator(Type &value, Extra &&...extra) {
-    return make_iterator<Policy>(std::begin(value), std::end(value), extra...);
+typing::Iterator<ValueType> make_iterator(Type &value, Extra &&...extra) {
+    return make_iterator<Policy>(
+        std::begin(value), std::end(value), std::forward<Extra>(extra)...);
 }
 
 /// Makes an iterator over the keys (`.first`) of a stl map-like container supporting
 /// `std::begin()`/`std::end()`
 template <return_value_policy Policy = return_value_policy::reference_internal,
           typename Type,
+          typename KeyType = typename detail::iterator_key_access<
+              decltype(std::begin(std::declval<Type &>()))>::result_type,
           typename... Extra>
-iterator make_key_iterator(Type &value, Extra &&...extra) {
-    return make_key_iterator<Policy>(std::begin(value), std::end(value), extra...);
+typing::Iterator<KeyType> make_key_iterator(Type &value, Extra &&...extra) {
+    return make_key_iterator<Policy>(
+        std::begin(value), std::end(value), std::forward<Extra>(extra)...);
 }
 
 /// Makes an iterator over the values (`.second`) of a stl map-like container supporting
 /// `std::begin()`/`std::end()`
 template <return_value_policy Policy = return_value_policy::reference_internal,
           typename Type,
+          typename ValueType = typename detail::iterator_value_access<
+              decltype(std::begin(std::declval<Type &>()))>::result_type,
           typename... Extra>
-iterator make_value_iterator(Type &value, Extra &&...extra) {
-    return make_value_iterator<Policy>(std::begin(value), std::end(value), extra...);
+typing::Iterator<ValueType> make_value_iterator(Type &value, Extra &&...extra) {
+    return make_value_iterator<Policy>(
+        std::begin(value), std::end(value), std::forward<Extra>(extra)...);
 }
 
 template <typename InputType, typename OutputType>
@@ -2431,7 +2595,11 @@ void implicitly_convertible() {
         ~set_flag() { flag = false; }
     };
     auto implicit_caster = [](PyObject *obj, PyTypeObject *type) -> PyObject * {
+#ifdef Py_GIL_DISABLED
+        thread_local bool currently_used = false;
+#else
         static bool currently_used = false;
+#endif
         if (currently_used) { // implicit conversions are non-reentrant
             return nullptr;
         }
@@ -2449,15 +2617,17 @@ void implicitly_convertible() {
     };
 
     if (auto *tinfo = detail::get_type_info(typeid(OutputType))) {
-        tinfo->implicit_conversions.push_back(implicit_caster);
+        tinfo->implicit_conversions.emplace_back(std::move(implicit_caster));
     } else {
         pybind11_fail("implicitly_convertible: Unable to find type " + type_id<OutputType>());
     }
 }
 
 inline void register_exception_translator(ExceptionTranslator &&translator) {
-    detail::get_internals().registered_exception_translators.push_front(
-        std::forward<ExceptionTranslator>(translator));
+    detail::with_internals([&](detail::internals &internals) {
+        internals.registered_exception_translators.push_front(
+            std::forward<ExceptionTranslator>(translator));
+    });
 }
 
 /**
@@ -2467,14 +2637,17 @@ inline void register_exception_translator(ExceptionTranslator &&translator) {
  * the exception.
  */
 inline void register_local_exception_translator(ExceptionTranslator &&translator) {
-    detail::get_local_internals().registered_exception_translators.push_front(
-        std::forward<ExceptionTranslator>(translator));
+    detail::with_internals([&](detail::internals &internals) {
+        (void) internals;
+        detail::get_local_internals().registered_exception_translators.push_front(
+            std::forward<ExceptionTranslator>(translator));
+    });
 }
 
 /**
  * Wrapper to generate a new Python exception type.
  *
- * This should only be used with PyErr_SetString for now.
+ * This should only be used with py::set_error() for now.
  * It is not (yet) possible to use as a py::base.
  * Template type argument is reserved for future use.
  */
@@ -2485,7 +2658,7 @@ class exception : public object {
     exception(handle scope, const char *name, handle base = PyExc_Exception) {
         std::string full_name
             = scope.attr("__name__").cast<std::string>() + std::string(".") + name;
-        m_ptr = PyErr_NewException(const_cast<char *>(full_name.c_str()), base.ptr(), NULL);
+        m_ptr = PyErr_NewException(const_cast<char *>(full_name.c_str()), base.ptr(), nullptr);
         if (hasattr(scope, "__dict__") && scope.attr("__dict__").contains(name)) {
             pybind11_fail("Error during initialization: multiple incompatible "
                           "definitions with name \""
@@ -2495,27 +2668,25 @@ class exception : public object {
     }
 
     // Sets the current python exception to this exception object with the given message
-    void operator()(const char *message) { PyErr_SetString(m_ptr, message); }
+    PYBIND11_DEPRECATED("Please use py::set_error() instead "
+                        "(https://github.com/pybind/pybind11/pull/4772)")
+    void operator()(const char *message) const { set_error(*this, message); }
 };
 
 PYBIND11_NAMESPACE_BEGIN(detail)
-// Returns a reference to a function-local static exception object used in the simple
-// register_exception approach below.  (It would be simpler to have the static local variable
-// directly in register_exception, but that makes clang <3.5 segfault - issue #1349).
-template <typename CppException>
-exception<CppException> &get_exception_object() {
-    static exception<CppException> ex;
-    return ex;
-}
+
+template <>
+struct handle_type_name<exception<void>> {
+    static constexpr auto name = const_name("Exception");
+};
 
 // Helper function for register_exception and register_local_exception
 template <typename CppException>
 exception<CppException> &
 register_exception_impl(handle scope, const char *name, handle base, bool isLocal) {
-    auto &ex = detail::get_exception_object<CppException>();
-    if (!ex) {
-        ex = exception<CppException>(scope, name, base);
-    }
+    PYBIND11_CONSTINIT static gil_safe_call_once_and_store<exception<CppException>> exc_storage;
+    exc_storage.call_once_and_store_result(
+        [&]() { return exception<CppException>(scope, name, base); });
 
     auto register_func
         = isLocal ? &register_local_exception_translator : &register_exception_translator;
@@ -2527,10 +2698,10 @@ register_exception_impl(handle scope, const char *name, handle base, bool isLoca
         try {
             std::rethrow_exception(p);
         } catch (const CppException &e) {
-            detail::get_exception_object<CppException>()(e.what());
+            set_error(exc_storage.get_stored(), e.what());
         }
     });
-    return ex;
+    return exc_storage.get_stored();
 }
 
 PYBIND11_NAMESPACE_END(detail)
@@ -2567,8 +2738,8 @@ PYBIND11_NOINLINE void print(const tuple &args, const dict &kwargs) {
     for (size_t i = 0; i < args.size(); ++i) {
         strings[i] = str(args[i]);
     }
-    auto sep = kwargs.contains("sep") ? kwargs["sep"] : cast(" ");
-    auto line = sep.attr("join")(strings);
+    auto sep = kwargs.contains("sep") ? kwargs["sep"] : str(" ");
+    auto line = sep.attr("join")(std::move(strings));
 
     object file;
     if (kwargs.contains("file")) {
@@ -2586,8 +2757,8 @@ PYBIND11_NOINLINE void print(const tuple &args, const dict &kwargs) {
     }
 
     auto write = file.attr("write");
-    write(line);
-    write(kwargs.contains("end") ? kwargs["end"] : cast("\n"));
+    write(std::move(line));
+    write(kwargs.contains("end") ? kwargs["end"] : str("\n"));
 
     if (kwargs.contains("flush") && kwargs["flush"].cast<bool>()) {
         file.attr("flush")();
@@ -2601,17 +2772,21 @@ void print(Args &&...args) {
     detail::print(c.args(), c.kwargs());
 }
 
-error_already_set::~error_already_set() {
-    if (m_type) {
-        gil_scoped_acquire gil;
-        error_scope scope;
-        m_type.release().dec_ref();
-        m_value.release().dec_ref();
-        m_trace.release().dec_ref();
-    }
+inline void
+error_already_set::m_fetched_error_deleter(detail::error_fetch_and_normalize *raw_ptr) {
+    gil_scoped_acquire gil;
+    error_scope scope;
+    delete raw_ptr;
+}
+
+inline const char *error_already_set::what() const noexcept {
+    gil_scoped_acquire gil;
+    error_scope scope;
+    return m_fetched_error->error_string().c_str();
 }
 
 PYBIND11_NAMESPACE_BEGIN(detail)
+
 inline function
 get_type_override(const void *this_ptr, const type_info *this_type, const char *name) {
     handle self = get_object_handle(this_ptr, this_type);
@@ -2623,14 +2798,19 @@ get_type_override(const void *this_ptr, const type_info *this_type, const char *
 
     /* Cache functions that aren't overridden in Python to avoid
        many costly Python dictionary lookups below */
-    auto &cache = get_internals().inactive_override_cache;
-    if (cache.find(key) != cache.end()) {
+    bool not_overridden = with_internals([&key](internals &internals) {
+        auto &cache = internals.inactive_override_cache;
+        return cache.find(key) != cache.end();
+    });
+    if (not_overridden) {
         return function();
     }
 
     function override = getattr(self, name, function());
     if (override.is_cpp_function()) {
-        cache.insert(key);
+        with_internals([&](internals &internals) {
+            internals.inactive_override_cache.insert(std::move(key));
+        });
         return function();
     }
 
@@ -2643,12 +2823,22 @@ get_type_override(const void *this_ptr, const type_info *this_type, const char *
         PyCodeObject *f_code = PyFrame_GetCode(frame);
         // f_code is guaranteed to not be NULL
         if ((std::string) str(f_code->co_name) == name && f_code->co_argcount > 0) {
+#        if PY_VERSION_HEX >= 0x030d0000
+            PyObject *locals = PyEval_GetFrameLocals();
+#        else
             PyObject *locals = PyEval_GetLocals();
+            Py_XINCREF(locals);
+#        endif
             if (locals != nullptr) {
+#        if PY_VERSION_HEX >= 0x030b0000
+                PyObject *co_varnames = PyCode_GetVarnames(f_code);
+#        else
                 PyObject *co_varnames = PyObject_GetAttrString((PyObject *) f_code, "co_varnames");
+#        endif
                 PyObject *self_arg = PyTuple_GET_ITEM(co_varnames, 0);
                 Py_DECREF(co_varnames);
                 PyObject *self_caller = dict_getitem(locals, self_arg);
+                Py_DECREF(locals);
                 if (self_caller == self.ptr()) {
                     Py_DECREF(f_code);
                     Py_DECREF(frame);
@@ -2725,10 +2915,14 @@ function get_override(const T *this_ptr, const char *name) {
             = pybind11::get_override(static_cast<const cname *>(this), name);                     \
         if (override) {                                                                           \
             auto o = override(__VA_ARGS__);                                                       \
-            if (pybind11::detail::cast_is_temporary_value_reference<ret_type>::value) {           \
+            PYBIND11_WARNING_PUSH                                                                 \
+            PYBIND11_WARNING_DISABLE_MSVC(4127)                                                   \
+            if (pybind11::detail::cast_is_temporary_value_reference<ret_type>::value              \
+                && !pybind11::detail::is_same_ignoring_cvref<ret_type, PyObject *>::value) {      \
                 static pybind11::detail::override_caster_t<ret_type> caster;                      \
                 return pybind11::detail::cast_ref<ret_type>(std::move(o), caster);                \
             }                                                                                     \
+            PYBIND11_WARNING_POP                                                                  \
             return pybind11::detail::cast_safe<ret_type>(std::move(o));                           \
         }                                                                                         \
     } while (false)
@@ -2830,7 +3024,3 @@ inline function get_overload(const T *this_ptr, const char *name) {
     PYBIND11_OVERRIDE_PURE(PYBIND11_TYPE(ret_type), PYBIND11_TYPE(cname), fn, __VA_ARGS__);
 
 PYBIND11_NAMESPACE_END(PYBIND11_NAMESPACE)
-
-#if defined(__GNUC__) && __GNUC__ == 7
-#    pragma GCC diagnostic pop // -Wnoexcept-type
-#endif
diff --git a/include/pybind11/pytypes.h b/include/pybind11/pytypes.h
index 18cd715805..f26c307a87 100644
--- a/include/pybind11/pytypes.h
+++ b/include/pybind11/pytypes.h
@@ -12,7 +12,15 @@
 #include "detail/common.h"
 #include "buffer_info.h"
 
+#include <assert.h>
+#include <cstddef>
+#include <exception>
+#include <frameobject.h>
+#include <iterator>
+#include <memory>
+#include <string>
 #include <type_traits>
+#include <typeinfo>
 #include <utility>
 
 #if defined(PYBIND11_HAS_OPTIONAL)
@@ -25,6 +33,8 @@
 
 PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)
 
+PYBIND11_WARNING_DISABLE_MSVC(4127)
+
 /* A few forward declarations */
 class handle;
 class object;
@@ -49,6 +59,7 @@ struct sequence_item;
 struct list_item;
 struct tuple_item;
 } // namespace accessor_policies
+// PLEASE KEEP handle_type_name SPECIALIZATIONS IN SYNC.
 using obj_attr_accessor = accessor<accessor_policies::obj_attr>;
 using str_attr_accessor = accessor<accessor_policies::str_attr>;
 using item_accessor = accessor<accessor_policies::generic_item>;
@@ -85,7 +96,9 @@ class object_api : public pyobject_tag {
         or `object` subclass causes a call to ``__setitem__``.
     \endrst */
     item_accessor operator[](handle key) const;
-    /// See above (the only difference is that they key is provided as a string literal)
+    /// See above (the only difference is that the key's reference is stolen)
+    item_accessor operator[](object &&key) const;
+    /// See above (the only difference is that the key is provided as a string literal)
     item_accessor operator[](const char *key) const;
 
     /** \rst
@@ -95,7 +108,9 @@ class object_api : public pyobject_tag {
         or `object` subclass causes a call to ``setattr``.
     \endrst */
     obj_attr_accessor attr(handle key) const;
-    /// See above (the only difference is that they key is provided as a string literal)
+    /// See above (the only difference is that the key's reference is stolen)
+    obj_attr_accessor attr(object &&key) const;
+    /// See above (the only difference is that the key is provided as a string literal)
     str_attr_accessor attr(const char *key) const;
 
     /** \rst
@@ -143,23 +158,23 @@ class object_api : public pyobject_tag {
     object operator-() const;
     object operator~() const;
     object operator+(object_api const &other) const;
-    object operator+=(object_api const &other) const;
+    object operator+=(object_api const &other);
     object operator-(object_api const &other) const;
-    object operator-=(object_api const &other) const;
+    object operator-=(object_api const &other);
     object operator*(object_api const &other) const;
-    object operator*=(object_api const &other) const;
+    object operator*=(object_api const &other);
     object operator/(object_api const &other) const;
-    object operator/=(object_api const &other) const;
+    object operator/=(object_api const &other);
     object operator|(object_api const &other) const;
-    object operator|=(object_api const &other) const;
+    object operator|=(object_api const &other);
     object operator&(object_api const &other) const;
-    object operator&=(object_api const &other) const;
+    object operator&=(object_api const &other);
     object operator^(object_api const &other) const;
-    object operator^=(object_api const &other) const;
+    object operator^=(object_api const &other);
     object operator<<(object_api const &other) const;
-    object operator<<=(object_api const &other) const;
+    object operator<<=(object_api const &other);
     object operator>>(object_api const &other) const;
-    object operator>>=(object_api const &other) const;
+    object operator>>=(object_api const &other);
 
     PYBIND11_DEPRECATED("Use py::str(obj) instead")
     pybind11::str str() const;
@@ -168,7 +183,15 @@ class object_api : public pyobject_tag {
     str_attr_accessor doc() const;
 
     /// Return the object's current reference count
-    int ref_count() const { return static_cast<int>(Py_REFCNT(derived().ptr())); }
+    ssize_t ref_count() const {
+#ifdef PYPY_VERSION
+        // PyPy uses the top few bits for REFCNT_FROM_PYPY & REFCNT_FROM_PYPY_LIGHT
+        // Following pybind11 2.12.1 and older behavior and removing this part
+        return static_cast<ssize_t>(static_cast<int>(Py_REFCNT(derived().ptr())));
+#else
+        return Py_REFCNT(derived().ptr());
+#endif
+    }
 
     // TODO PYBIND11_DEPRECATED(
     //     "Call py::type::handle_of(h) or py::type::of(h) instead of h.get_type()")
@@ -178,8 +201,17 @@ class object_api : public pyobject_tag {
     bool rich_compare(object_api const &other, int value) const;
 };
 
+template <typename T>
+using is_pyobj_ptr_or_nullptr_t = detail::any_of<std::is_same<T, PyObject *>,
+                                                 std::is_same<T, PyObject *const>,
+                                                 std::is_same<T, std::nullptr_t>>;
+
 PYBIND11_NAMESPACE_END(detail)
 
+#if !defined(PYBIND11_HANDLE_REF_DEBUG) && !defined(NDEBUG)
+#    define PYBIND11_HANDLE_REF_DEBUG
+#endif
+
 /** \rst
     Holds a reference to a Python object (no reference counting)
 
@@ -195,9 +227,24 @@ class handle : public detail::object_api<handle> {
 public:
     /// The default constructor creates a handle with a ``nullptr``-valued pointer
     handle() = default;
-    /// Creates a ``handle`` from the given raw Python object pointer
+
+    /// Enable implicit conversion from ``PyObject *`` and ``nullptr``.
+    /// Not using ``handle(PyObject *ptr)`` to avoid implicit conversion from ``0``.
+    template <typename T,
+              detail::enable_if_t<detail::is_pyobj_ptr_or_nullptr_t<T>::value, int> = 0>
     // NOLINTNEXTLINE(google-explicit-constructor)
-    handle(PyObject *ptr) : m_ptr(ptr) {} // Allow implicit conversion from PyObject*
+    handle(T ptr) : m_ptr(ptr) {}
+
+    /// Enable implicit conversion through ``T::operator PyObject *()``.
+    template <
+        typename T,
+        detail::enable_if_t<detail::all_of<detail::none_of<std::is_base_of<handle, T>,
+                                                           detail::is_pyobj_ptr_or_nullptr_t<T>>,
+                                           std::is_convertible<T, PyObject *>>::value,
+                            int>
+        = 0>
+    // NOLINTNEXTLINE(google-explicit-constructor)
+    handle(T &obj) : m_ptr(obj) {}
 
     /// Return the underlying ``PyObject *`` pointer
     PyObject *ptr() const { return m_ptr; }
@@ -209,6 +256,14 @@ class handle : public detail::object_api<handle> {
         this function automatically. Returns a reference to itself.
     \endrst */
     const handle &inc_ref() const & {
+#ifdef PYBIND11_HANDLE_REF_DEBUG
+        inc_ref_counter(1);
+#endif
+#ifdef PYBIND11_ASSERT_GIL_HELD_INCREF_DECREF
+        if (m_ptr != nullptr && !PyGILState_Check()) {
+            throw_gilstate_error("pybind11::handle::inc_ref()");
+        }
+#endif
         Py_XINCREF(m_ptr);
         return *this;
     }
@@ -219,6 +274,11 @@ class handle : public detail::object_api<handle> {
         this function automatically. Returns a reference to itself.
     \endrst */
     const handle &dec_ref() const & {
+#ifdef PYBIND11_ASSERT_GIL_HELD_INCREF_DECREF
+        if (m_ptr != nullptr && !PyGILState_Check()) {
+            throw_gilstate_error("pybind11::handle::dec_ref()");
+        }
+#endif
         Py_XDECREF(m_ptr);
         return *this;
     }
@@ -244,8 +304,53 @@ class handle : public detail::object_api<handle> {
 
 protected:
     PyObject *m_ptr = nullptr;
+
+private:
+#ifdef PYBIND11_ASSERT_GIL_HELD_INCREF_DECREF
+    void throw_gilstate_error(const std::string &function_name) const {
+        fprintf(
+            stderr,
+            "%s is being called while the GIL is either not held or invalid. Please see "
+            "https://pybind11.readthedocs.io/en/stable/advanced/"
+            "misc.html#common-sources-of-global-interpreter-lock-errors for debugging advice.\n"
+            "If you are convinced there is no bug in your code, you can #define "
+            "PYBIND11_NO_ASSERT_GIL_HELD_INCREF_DECREF "
+            "to disable this check. In that case you have to ensure this #define is consistently "
+            "used for all translation units linked into a given pybind11 extension, otherwise "
+            "there will be ODR violations.",
+            function_name.c_str());
+        if (Py_TYPE(m_ptr)->tp_name != nullptr) {
+            fprintf(stderr,
+                    " The failing %s call was triggered on a %s object.",
+                    function_name.c_str(),
+                    Py_TYPE(m_ptr)->tp_name);
+        }
+        fprintf(stderr, "\n");
+        fflush(stderr);
+        throw std::runtime_error(function_name + " PyGILState_Check() failure.");
+    }
+#endif
+
+#ifdef PYBIND11_HANDLE_REF_DEBUG
+    static std::size_t inc_ref_counter(std::size_t add) {
+        thread_local std::size_t counter = 0;
+        counter += add;
+        return counter;
+    }
+
+public:
+    static std::size_t inc_ref_counter() { return inc_ref_counter(0); }
+#endif
 };
 
+inline void set_error(const handle &type, const char *message) {
+    PyErr_SetString(type.ptr(), message);
+}
+
+inline void set_error(const handle &type, const handle &value) {
+    PyErr_SetObject(type.ptr(), value.ptr());
+}
+
 /** \rst
     Holds a reference to a Python object (with reference counting)
 
@@ -268,10 +373,7 @@ class object : public handle {
     /// Copy constructor; always increases the reference count
     object(const object &o) : handle(o) { inc_ref(); }
     /// Move constructor; steals the object from ``other`` and preserves its reference count
-    object(object &&other) noexcept {
-        m_ptr = other.m_ptr;
-        other.m_ptr = nullptr;
-    }
+    object(object &&other) noexcept : handle(other) { other.m_ptr = nullptr; }
     /// Destructor; automatically calls `handle::dec_ref()`
     ~object() { dec_ref(); }
 
@@ -287,12 +389,15 @@ class object : public handle {
     }
 
     object &operator=(const object &other) {
-        other.inc_ref();
-        // Use temporary variable to ensure `*this` remains valid while
-        // `Py_XDECREF` executes, in case `*this` is accessible from Python.
-        handle temp(m_ptr);
-        m_ptr = other.m_ptr;
-        temp.dec_ref();
+        // Skip inc_ref and dec_ref if both objects are the same
+        if (!this->is(other)) {
+            other.inc_ref();
+            // Use temporary variable to ensure `*this` remains valid while
+            // `Py_XDECREF` executes, in case `*this` is accessible from Python.
+            handle temp(m_ptr);
+            m_ptr = other.m_ptr;
+            temp.dec_ref();
+        }
         return *this;
     }
 
@@ -306,6 +411,20 @@ class object : public handle {
         return *this;
     }
 
+#define PYBIND11_INPLACE_OP(iop)                                                                  \
+    object iop(object_api const &other) { return operator=(handle::iop(other)); }
+
+    PYBIND11_INPLACE_OP(operator+=)
+    PYBIND11_INPLACE_OP(operator-=)
+    PYBIND11_INPLACE_OP(operator*=)
+    PYBIND11_INPLACE_OP(operator/=)
+    PYBIND11_INPLACE_OP(operator|=)
+    PYBIND11_INPLACE_OP(operator&=)
+    PYBIND11_INPLACE_OP(operator^=)
+    PYBIND11_INPLACE_OP(operator<<=)
+    PYBIND11_INPLACE_OP(operator>>=)
+#undef PYBIND11_INPLACE_OP
+
     // Calling cast() on an object lvalue just copies (via handle::cast)
     template <typename T>
     T cast() const &;
@@ -363,51 +482,291 @@ T reinterpret_steal(handle h) {
 }
 
 PYBIND11_NAMESPACE_BEGIN(detail)
+
+// Equivalent to obj.__class__.__name__ (or obj.__name__ if obj is a class).
+inline const char *obj_class_name(PyObject *obj) {
+    if (PyType_Check(obj)) {
+        return reinterpret_cast<PyTypeObject *>(obj)->tp_name;
+    }
+    return Py_TYPE(obj)->tp_name;
+}
+
 std::string error_string();
-PYBIND11_NAMESPACE_END(detail)
 
-#if defined(_MSC_VER)
-#    pragma warning(push)
-#    pragma warning(disable : 4275 4251)
-//     warning C4275: An exported class was derived from a class that wasn't exported.
-//     Can be ignored when derived from a STL class.
-#endif
-/// Fetch and hold an error which was already set in Python.  An instance of this is typically
-/// thrown to propagate python-side errors back through C++ which can either be caught manually or
-/// else falls back to the function dispatcher (which then raises the captured error back to
-/// python).
-class PYBIND11_EXPORT_EXCEPTION error_already_set : public std::runtime_error {
-public:
-    /// Constructs a new exception from the current Python error indicator, if any.  The current
-    /// Python error indicator will be cleared.
-    error_already_set() : std::runtime_error(detail::error_string()) {
+// The code in this struct is very unusual, to minimize the chances of
+// masking bugs (elsewhere) by errors during the error handling (here).
+// This is meant to be a lifeline for troubleshooting long-running processes
+// that crash under conditions that are virtually impossible to reproduce.
+// Low-level implementation alternatives are preferred to higher-level ones
+// that might raise cascading exceptions. Last-ditch-kind-of attempts are made
+// to report as much of the original error as possible, even if there are
+// secondary issues obtaining some of the details.
+struct error_fetch_and_normalize {
+    // This comment only applies to Python <= 3.11:
+    //     Immediate normalization is long-established behavior (starting with
+    //     https://github.com/pybind/pybind11/commit/135ba8deafb8bf64a15b24d1513899eb600e2011
+    //     from Sep 2016) and safest. Normalization could be deferred, but this could mask
+    //     errors elsewhere, the performance gain is very minor in typical situations
+    //     (usually the dominant bottleneck is EH unwinding), and the implementation here
+    //     would be more complex.
+    // Starting with Python 3.12, PyErr_Fetch() normalizes exceptions immediately.
+    // Any errors during normalization are tracked under __notes__.
+    explicit error_fetch_and_normalize(const char *called) {
         PyErr_Fetch(&m_type.ptr(), &m_value.ptr(), &m_trace.ptr());
+        if (!m_type) {
+            pybind11_fail("Internal error: " + std::string(called)
+                          + " called while "
+                            "Python error indicator not set.");
+        }
+        const char *exc_type_name_orig = detail::obj_class_name(m_type.ptr());
+        if (exc_type_name_orig == nullptr) {
+            pybind11_fail("Internal error: " + std::string(called)
+                          + " failed to obtain the name "
+                            "of the original active exception type.");
+        }
+        m_lazy_error_string = exc_type_name_orig;
+#if PY_VERSION_HEX >= 0x030C0000
+        // The presence of __notes__ is likely due to exception normalization
+        // errors, although that is not necessarily true, therefore insert a
+        // hint only:
+        if (PyObject_HasAttrString(m_value.ptr(), "__notes__")) {
+            m_lazy_error_string += "[WITH __notes__]";
+        }
+#else
+        // PyErr_NormalizeException() may change the exception type if there are cascading
+        // failures. This can potentially be extremely confusing.
+        PyErr_NormalizeException(&m_type.ptr(), &m_value.ptr(), &m_trace.ptr());
+        if (m_type.ptr() == nullptr) {
+            pybind11_fail("Internal error: " + std::string(called)
+                          + " failed to normalize the "
+                            "active exception.");
+        }
+        const char *exc_type_name_norm = detail::obj_class_name(m_type.ptr());
+        if (exc_type_name_norm == nullptr) {
+            pybind11_fail("Internal error: " + std::string(called)
+                          + " failed to obtain the name "
+                            "of the normalized active exception type.");
+        }
+#    if defined(PYPY_VERSION_NUM) && PYPY_VERSION_NUM < 0x07030a00
+        // This behavior runs the risk of masking errors in the error handling, but avoids a
+        // conflict with PyPy, which relies on the normalization here to change OSError to
+        // FileNotFoundError (https://github.com/pybind/pybind11/issues/4075).
+        m_lazy_error_string = exc_type_name_norm;
+#    else
+        if (exc_type_name_norm != m_lazy_error_string) {
+            std::string msg = std::string(called)
+                              + ": MISMATCH of original and normalized "
+                                "active exception types: ";
+            msg += "ORIGINAL ";
+            msg += m_lazy_error_string;
+            msg += " REPLACED BY ";
+            msg += exc_type_name_norm;
+            msg += ": " + format_value_and_trace();
+            pybind11_fail(msg);
+        }
+#    endif
+#endif
     }
 
-    error_already_set(const error_already_set &) = default;
-    error_already_set(error_already_set &&) = default;
+    error_fetch_and_normalize(const error_fetch_and_normalize &) = delete;
+    error_fetch_and_normalize(error_fetch_and_normalize &&) = delete;
+
+    std::string format_value_and_trace() const {
+        std::string result;
+        std::string message_error_string;
+        if (m_value) {
+            auto value_str = reinterpret_steal<object>(PyObject_Str(m_value.ptr()));
+            constexpr const char *message_unavailable_exc
+                = "<MESSAGE UNAVAILABLE DUE TO ANOTHER EXCEPTION>";
+            if (!value_str) {
+                message_error_string = detail::error_string();
+                result = message_unavailable_exc;
+            } else {
+                // Not using `value_str.cast<std::string>()`, to not potentially throw a secondary
+                // error_already_set that will then result in process termination (#4288).
+                auto value_bytes = reinterpret_steal<object>(
+                    PyUnicode_AsEncodedString(value_str.ptr(), "utf-8", "backslashreplace"));
+                if (!value_bytes) {
+                    message_error_string = detail::error_string();
+                    result = message_unavailable_exc;
+                } else {
+                    char *buffer = nullptr;
+                    Py_ssize_t length = 0;
+                    if (PyBytes_AsStringAndSize(value_bytes.ptr(), &buffer, &length) == -1) {
+                        message_error_string = detail::error_string();
+                        result = message_unavailable_exc;
+                    } else {
+                        result = std::string(buffer, static_cast<std::size_t>(length));
+                    }
+                }
+            }
+#if PY_VERSION_HEX >= 0x030B0000
+            auto notes
+                = reinterpret_steal<object>(PyObject_GetAttrString(m_value.ptr(), "__notes__"));
+            if (!notes) {
+                PyErr_Clear(); // No notes is good news.
+            } else {
+                auto len_notes = PyList_Size(notes.ptr());
+                if (len_notes < 0) {
+                    result += "\nFAILURE obtaining len(__notes__): " + detail::error_string();
+                } else {
+                    result += "\n__notes__ (len=" + std::to_string(len_notes) + "):";
+                    for (ssize_t i = 0; i < len_notes; i++) {
+                        PyObject *note = PyList_GET_ITEM(notes.ptr(), i);
+                        auto note_bytes = reinterpret_steal<object>(
+                            PyUnicode_AsEncodedString(note, "utf-8", "backslashreplace"));
+                        if (!note_bytes) {
+                            result += "\nFAILURE obtaining __notes__[" + std::to_string(i)
+                                      + "]: " + detail::error_string();
+                        } else {
+                            char *buffer = nullptr;
+                            Py_ssize_t length = 0;
+                            if (PyBytes_AsStringAndSize(note_bytes.ptr(), &buffer, &length)
+                                == -1) {
+                                result += "\nFAILURE formatting __notes__[" + std::to_string(i)
+                                          + "]: " + detail::error_string();
+                            } else {
+                                result += '\n';
+                                result += std::string(buffer, static_cast<std::size_t>(length));
+                            }
+                        }
+                    }
+                }
+            }
+#endif
+        } else {
+            result = "<MESSAGE UNAVAILABLE>";
+        }
+        if (result.empty()) {
+            result = "<EMPTY MESSAGE>";
+        }
+
+        bool have_trace = false;
+        if (m_trace) {
+#if !defined(PYPY_VERSION)
+            auto *tb = reinterpret_cast<PyTracebackObject *>(m_trace.ptr());
+
+            // Get the deepest trace possible.
+            while (tb->tb_next) {
+                tb = tb->tb_next;
+            }
+
+            PyFrameObject *frame = tb->tb_frame;
+            Py_XINCREF(frame);
+            result += "\n\nAt:\n";
+            while (frame) {
+#    if PY_VERSION_HEX >= 0x030900B1
+                PyCodeObject *f_code = PyFrame_GetCode(frame);
+#    else
+                PyCodeObject *f_code = frame->f_code;
+                Py_INCREF(f_code);
+#    endif
+                int lineno = PyFrame_GetLineNumber(frame);
+                result += "  ";
+                result += handle(f_code->co_filename).cast<std::string>();
+                result += '(';
+                result += std::to_string(lineno);
+                result += "): ";
+                result += handle(f_code->co_name).cast<std::string>();
+                result += '\n';
+                Py_DECREF(f_code);
+#    if PY_VERSION_HEX >= 0x030900B1
+                auto *b_frame = PyFrame_GetBack(frame);
+#    else
+                auto *b_frame = frame->f_back;
+                Py_XINCREF(b_frame);
+#    endif
+                Py_DECREF(frame);
+                frame = b_frame;
+            }
+
+            have_trace = true;
+#endif //! defined(PYPY_VERSION)
+        }
+
+        if (!message_error_string.empty()) {
+            if (!have_trace) {
+                result += '\n';
+            }
+            result += "\nMESSAGE UNAVAILABLE DUE TO EXCEPTION: " + message_error_string;
+        }
+
+        return result;
+    }
 
-    inline ~error_already_set() override;
+    std::string const &error_string() const {
+        if (!m_lazy_error_string_completed) {
+            m_lazy_error_string += ": " + format_value_and_trace();
+            m_lazy_error_string_completed = true;
+        }
+        return m_lazy_error_string;
+    }
 
-    /// Give the currently-held error back to Python, if any.  If there is currently a Python error
-    /// already set it is cleared first.  After this call, the current object no longer stores the
-    /// error variables (but the `.what()` string is still available).
     void restore() {
-        PyErr_Restore(m_type.release().ptr(), m_value.release().ptr(), m_trace.release().ptr());
+        if (m_restore_called) {
+            pybind11_fail("Internal error: pybind11::detail::error_fetch_and_normalize::restore() "
+                          "called a second time. ORIGINAL ERROR: "
+                          + error_string());
+        }
+        PyErr_Restore(m_type.inc_ref().ptr(), m_value.inc_ref().ptr(), m_trace.inc_ref().ptr());
+        m_restore_called = true;
+    }
+
+    bool matches(handle exc) const {
+        return (PyErr_GivenExceptionMatches(m_type.ptr(), exc.ptr()) != 0);
     }
 
+    // Not protecting these for simplicity.
+    object m_type, m_value, m_trace;
+
+private:
+    // Only protecting invariants.
+    mutable std::string m_lazy_error_string;
+    mutable bool m_lazy_error_string_completed = false;
+    mutable bool m_restore_called = false;
+};
+
+inline std::string error_string() {
+    return error_fetch_and_normalize("pybind11::detail::error_string").error_string();
+}
+
+PYBIND11_NAMESPACE_END(detail)
+
+/// Fetch and hold an error which was already set in Python.  An instance of this is typically
+/// thrown to propagate python-side errors back through C++ which can either be caught manually or
+/// else falls back to the function dispatcher (which then raises the captured error back to
+/// python).
+class PYBIND11_EXPORT_EXCEPTION error_already_set : public std::exception {
+public:
+    /// Fetches the current Python exception (using PyErr_Fetch()), which will clear the
+    /// current Python error indicator.
+    error_already_set()
+        : m_fetched_error{new detail::error_fetch_and_normalize("pybind11::error_already_set"),
+                          m_fetched_error_deleter} {}
+
+    /// The what() result is built lazily on demand.
+    /// WARNING: This member function needs to acquire the Python GIL. This can lead to
+    ///          crashes (undefined behavior) if the Python interpreter is finalizing.
+    const char *what() const noexcept override;
+
+    /// Restores the currently-held Python error (which will clear the Python error indicator first
+    /// if already set).
+    /// NOTE: This member function will always restore the normalized exception, which may or may
+    ///       not be the original Python exception.
+    /// WARNING: The GIL must be held when this member function is called!
+    void restore() { m_fetched_error->restore(); }
+
     /// If it is impossible to raise the currently-held error, such as in a destructor, we can
     /// write it out using Python's unraisable hook (`sys.unraisablehook`). The error context
     /// should be some object whose `repr()` helps identify the location of the error. Python
-    /// already knows the type and value of the error, so there is no need to repeat that. After
-    /// this call, the current object no longer stores the error variables, and neither does
-    /// Python.
+    /// already knows the type and value of the error, so there is no need to repeat that.
     void discard_as_unraisable(object err_context) {
         restore();
         PyErr_WriteUnraisable(err_context.ptr());
     }
     /// An alternate version of `discard_as_unraisable()`, where a string provides information on
     /// the location of the error. For example, `__func__` could be helpful.
+    /// WARNING: The GIL must be held when this member function is called!
     void discard_as_unraisable(const char *err_context) {
         discard_as_unraisable(reinterpret_steal<object>(PYBIND11_FROM_STRING(err_context)));
     }
@@ -419,20 +778,19 @@ class PYBIND11_EXPORT_EXCEPTION error_already_set : public std::runtime_error {
     /// Check if the currently trapped error type matches the given Python exception class (or a
     /// subclass thereof).  May also be passed a tuple to search for any exception class matches in
     /// the given tuple.
-    bool matches(handle exc) const {
-        return (PyErr_GivenExceptionMatches(m_type.ptr(), exc.ptr()) != 0);
-    }
+    bool matches(handle exc) const { return m_fetched_error->matches(exc); }
 
-    const object &type() const { return m_type; }
-    const object &value() const { return m_value; }
-    const object &trace() const { return m_trace; }
+    const object &type() const { return m_fetched_error->m_type; }
+    const object &value() const { return m_fetched_error->m_value; }
+    const object &trace() const { return m_fetched_error->m_trace; }
 
 private:
-    object m_type, m_value, m_trace;
+    std::shared_ptr<detail::error_fetch_and_normalize> m_fetched_error;
+
+    /// WARNING: This custom deleter needs to acquire the Python GIL. This can lead to
+    ///          crashes (undefined behavior) if the Python interpreter is finalizing.
+    static void m_fetched_error_deleter(detail::error_fetch_and_normalize *raw_ptr);
 };
-#if defined(_MSC_VER)
-#    pragma warning(pop)
-#endif
 
 /// Replaces the current Python error indicator with the chosen error, performing a
 /// 'raise from' to indicate that the chosen error was caused by the original error.
@@ -464,8 +822,7 @@ inline void raise_from(PyObject *type, const char *message) {
 
 /// Sets the current Python error indicator with the chosen error, performing a 'raise from'
 /// from the error contained in error_already_set to indicate that the chosen error was
-/// caused by the original error. After this function is called error_already_set will
-/// no longer contain an error.
+/// caused by the original error.
 inline void raise_from(error_already_set &err, PyObject *type, const char *message) {
     err.restore();
     raise_from(type, message);
@@ -603,13 +960,13 @@ inline handle get_function(handle value) {
 inline PyObject *dict_getitemstring(PyObject *v, const char *key) {
     PyObject *kv = nullptr, *rv = nullptr;
     kv = PyUnicode_FromString(key);
-    if (kv == NULL) {
+    if (kv == nullptr) {
         throw error_already_set();
     }
 
     rv = PyDict_GetItemWithError(v, kv);
     Py_DECREF(kv);
-    if (rv == NULL && PyErr_Occurred()) {
+    if (rv == nullptr && PyErr_Occurred()) {
         throw error_already_set();
     }
     return rv;
@@ -617,10 +974,27 @@ inline PyObject *dict_getitemstring(PyObject *v, const char *key) {
 
 inline PyObject *dict_getitem(PyObject *v, PyObject *key) {
     PyObject *rv = PyDict_GetItemWithError(v, key);
-    if (rv == NULL && PyErr_Occurred()) {
+    if (rv == nullptr && PyErr_Occurred()) {
+        throw error_already_set();
+    }
+    return rv;
+}
+
+inline PyObject *dict_getitemstringref(PyObject *v, const char *key) {
+#if PY_VERSION_HEX >= 0x030D0000
+    PyObject *rv;
+    if (PyDict_GetItemStringRef(v, key, &rv) < 0) {
         throw error_already_set();
     }
     return rv;
+#else
+    PyObject *rv = dict_getitemstring(v, key);
+    if (rv == nullptr && PyErr_Occurred()) {
+        throw error_already_set();
+    }
+    Py_XINCREF(rv);
+    return rv;
+#endif
 }
 
 // Helper aliases/functions to support implicit casting of values given to python
@@ -636,10 +1010,8 @@ object object_or_cast(T &&o);
 // Match a PyObject*, which we want to convert directly to handle via its converting constructor
 inline handle object_or_cast(PyObject *ptr) { return ptr; }
 
-#if defined(_MSC_VER) && _MSC_VER < 1920
-#    pragma warning(push)
-#    pragma warning(disable : 4522) // warning C4522: multiple assignment operators specified
-#endif
+PYBIND11_WARNING_PUSH
+PYBIND11_WARNING_DISABLE_MSVC(4522) // warning C4522: multiple assignment operators specified
 template <typename Policy>
 class accessor : public object_api<accessor<Policy>> {
     using key_type = typename Policy::key_type;
@@ -660,7 +1032,7 @@ class accessor : public object_api<accessor<Policy>> {
     }
     template <typename T>
     void operator=(T &&value) & {
-        get_cache() = reinterpret_borrow<object>(object_or_cast(std::forward<T>(value)));
+        get_cache() = ensure_object(object_or_cast(std::forward<T>(value)));
     }
 
     template <typename T = Policy>
@@ -688,6 +1060,9 @@ class accessor : public object_api<accessor<Policy>> {
     }
 
 private:
+    static object ensure_object(object &&o) { return std::move(o); }
+    static object ensure_object(handle h) { return reinterpret_borrow<object>(h); }
+
     object &get_cache() const {
         if (!cache) {
             cache = Policy::get(obj, key);
@@ -700,9 +1075,7 @@ class accessor : public object_api<accessor<Policy>> {
     key_type key;
     mutable object cache;
 };
-#if defined(_MSC_VER) && _MSC_VER < 1920
-#    pragma warning(pop)
-#endif
+PYBIND11_WARNING_POP
 
 PYBIND11_NAMESPACE_BEGIN(accessor_policies)
 struct obj_attr {
@@ -1048,7 +1421,7 @@ public:
 
 #define PYBIND11_OBJECT_CVT_DEFAULT(Name, Parent, CheckFun, ConvertFun)                           \
     PYBIND11_OBJECT_CVT(Name, Parent, CheckFun, ConvertFun)                                       \
-    Name() : Parent() {}
+    Name() = default;
 
 #define PYBIND11_OBJECT_CHECK_FAILED(Name, o_ptr)                                                 \
     ::pybind11::type_error("Object of type '"                                                     \
@@ -1071,7 +1444,7 @@ public:
 
 #define PYBIND11_OBJECT_DEFAULT(Name, Parent, CheckFun)                                           \
     PYBIND11_OBJECT(Name, Parent, CheckFun)                                                       \
-    Name() : Parent() {}
+    Name() = default;
 
 /// \addtogroup pytypes
 /// @{
@@ -1140,7 +1513,7 @@ class iterator : public object {
 private:
     void advance() {
         value = reinterpret_steal<object>(PyIter_Next(m_ptr));
-        if (PyErr_Occurred()) {
+        if (value.ptr() == nullptr && PyErr_Occurred()) {
             throw error_already_set();
         }
     }
@@ -1190,6 +1563,9 @@ class str : public object {
     str(const char *c, const SzType &n)
         : object(PyUnicode_FromStringAndSize(c, ssize_t_cast(n)), stolen_t{}) {
         if (!m_ptr) {
+            if (PyErr_Occurred()) {
+                throw error_already_set();
+            }
             pybind11_fail("Could not allocate string object!");
         }
     }
@@ -1199,6 +1575,9 @@ class str : public object {
     // NOLINTNEXTLINE(google-explicit-constructor)
     str(const char *c = "") : object(PyUnicode_FromString(c), stolen_t{}) {
         if (!m_ptr) {
+            if (PyErr_Occurred()) {
+                throw error_already_set();
+            }
             pybind11_fail("Could not allocate string object!");
         }
     }
@@ -1267,7 +1646,15 @@ inline namespace literals {
 /** \rst
     String literal version of `str`
  \endrst */
-inline str operator"" _s(const char *s, size_t size) { return {s, size}; }
+inline str
+#if !defined(__clang__) && defined(__GNUC__) && __GNUC__ < 5
+operator"" _s // gcc 4.8.5 insists on having a space (hard error).
+#else
+operator""_s // clang 17 generates a deprecation warning if there is a space.
+#endif
+    (const char *s, size_t size) {
+    return {s, size};
+}
 } // namespace literals
 
 /// \addtogroup pytypes
@@ -1356,6 +1743,9 @@ inline str::str(const bytes &b) {
     }
     auto obj = reinterpret_steal<object>(PyUnicode_FromStringAndSize(buffer, length));
     if (!obj) {
+        if (PyErr_Occurred()) {
+            throw error_already_set();
+        }
         pybind11_fail("Could not allocate string object!");
     }
     m_ptr = obj.release().ptr();
@@ -1433,7 +1823,7 @@ PYBIND11_NAMESPACE_BEGIN(detail)
 // unsigned type: (A)-1 != (B)-1 when A and B are unsigned types of different sizes).
 template <typename Unsigned>
 Unsigned as_unsigned(PyObject *o) {
-    if (PYBIND11_SILENCE_MSVC_C4127(sizeof(Unsigned) <= sizeof(unsigned long))) {
+    if (sizeof(Unsigned) <= sizeof(unsigned long)) {
         unsigned long v = PyLong_AsUnsignedLong(o);
         return v == (unsigned long) -1 && PyErr_Occurred() ? (Unsigned) -1 : (Unsigned) v;
     }
@@ -1450,7 +1840,7 @@ class int_ : public object {
     template <typename T, detail::enable_if_t<std::is_integral<T>::value, int> = 0>
     // NOLINTNEXTLINE(google-explicit-constructor)
     int_(T value) {
-        if (PYBIND11_SILENCE_MSVC_C4127(sizeof(T) <= sizeof(long))) {
+        if (sizeof(T) <= sizeof(long)) {
             if (std::is_signed<T>::value) {
                 m_ptr = PyLong_FromLong((long) value);
             } else {
@@ -1519,8 +1909,8 @@ class weakref : public object {
 class slice : public object {
 public:
     PYBIND11_OBJECT_DEFAULT(slice, object, PySlice_Check)
-    slice(handle start, handle stop, handle step) {
-        m_ptr = PySlice_New(start.ptr(), stop.ptr(), step.ptr());
+    slice(handle start, handle stop, handle step)
+        : object(PySlice_New(start.ptr(), stop.ptr(), step.ptr()), stolen_t{}) {
         if (!m_ptr) {
             pybind11_fail("Could not allocate slice object!");
         }
@@ -1567,45 +1957,34 @@ class capsule : public object {
 
     explicit capsule(const void *value,
                      const char *name = nullptr,
-                     void (*destructor)(PyObject *) = nullptr)
+                     PyCapsule_Destructor destructor = nullptr)
         : object(PyCapsule_New(const_cast<void *>(value), name, destructor), stolen_t{}) {
         if (!m_ptr) {
             throw error_already_set();
         }
     }
 
-    PYBIND11_DEPRECATED("Please pass a destructor that takes a void pointer as input")
-    capsule(const void *value, void (*destruct)(PyObject *))
-        : object(PyCapsule_New(const_cast<void *>(value), nullptr, destruct), stolen_t{}) {
+    PYBIND11_DEPRECATED("Please use the ctor with value, name, destructor args")
+    capsule(const void *value, PyCapsule_Destructor destructor)
+        : object(PyCapsule_New(const_cast<void *>(value), nullptr, destructor), stolen_t{}) {
         if (!m_ptr) {
             throw error_already_set();
         }
     }
 
+    /// Capsule name is nullptr.
     capsule(const void *value, void (*destructor)(void *)) {
-        m_ptr = PyCapsule_New(const_cast<void *>(value), nullptr, [](PyObject *o) {
-            auto destructor = reinterpret_cast<void (*)(void *)>(PyCapsule_GetContext(o));
-            if (destructor == nullptr) {
-                if (PyErr_Occurred()) {
-                    throw error_already_set();
-                }
-                pybind11_fail("Unable to get capsule context");
-            }
-            void *ptr = PyCapsule_GetPointer(o, nullptr);
-            if (ptr == nullptr) {
-                throw error_already_set();
-            }
-            destructor(ptr);
-        });
+        initialize_with_void_ptr_destructor(value, nullptr, destructor);
+    }
 
-        if (!m_ptr || PyCapsule_SetContext(m_ptr, (void *) destructor) != 0) {
-            throw error_already_set();
-        }
+    capsule(const void *value, const char *name, void (*destructor)(void *)) {
+        initialize_with_void_ptr_destructor(value, name, destructor);
     }
 
     explicit capsule(void (*destructor)()) {
         m_ptr = PyCapsule_New(reinterpret_cast<void *>(destructor), nullptr, [](PyObject *o) {
-            auto destructor = reinterpret_cast<void (*)()>(PyCapsule_GetPointer(o, nullptr));
+            const char *name = get_name_in_error_scope(o);
+            auto destructor = reinterpret_cast<void (*)()>(PyCapsule_GetPointer(o, name));
             if (destructor == nullptr) {
                 throw error_already_set();
             }
@@ -1640,7 +2019,59 @@ class capsule : public object {
         }
     }
 
-    const char *name() const { return PyCapsule_GetName(m_ptr); }
+    const char *name() const {
+        const char *name = PyCapsule_GetName(m_ptr);
+        if ((name == nullptr) && PyErr_Occurred()) {
+            throw error_already_set();
+        }
+        return name;
+    }
+
+    /// Replaces a capsule's name *without* calling the destructor on the existing one.
+    void set_name(const char *new_name) {
+        if (PyCapsule_SetName(m_ptr, new_name) != 0) {
+            throw error_already_set();
+        }
+    }
+
+private:
+    static const char *get_name_in_error_scope(PyObject *o) {
+        error_scope error_guard;
+
+        const char *name = PyCapsule_GetName(o);
+        if ((name == nullptr) && PyErr_Occurred()) {
+            // write out and consume error raised by call to PyCapsule_GetName
+            PyErr_WriteUnraisable(o);
+        }
+
+        return name;
+    }
+
+    void initialize_with_void_ptr_destructor(const void *value,
+                                             const char *name,
+                                             void (*destructor)(void *)) {
+        m_ptr = PyCapsule_New(const_cast<void *>(value), name, [](PyObject *o) {
+            // guard if destructor called while err indicator is set
+            error_scope error_guard;
+            auto destructor = reinterpret_cast<void (*)(void *)>(PyCapsule_GetContext(o));
+            if (destructor == nullptr && PyErr_Occurred()) {
+                throw error_already_set();
+            }
+            const char *name = get_name_in_error_scope(o);
+            void *ptr = PyCapsule_GetPointer(o, name);
+            if (ptr == nullptr) {
+                throw error_already_set();
+            }
+
+            if (destructor != nullptr) {
+                destructor(ptr);
+            }
+        });
+
+        if (!m_ptr || PyCapsule_SetContext(m_ptr, reinterpret_cast<void *>(destructor)) != 0) {
+            throw error_already_set();
+        }
+    }
 };
 
 class tuple : public object {
@@ -1657,7 +2088,10 @@ class tuple : public object {
     size_t size() const { return (size_t) PyTuple_Size(m_ptr); }
     bool empty() const { return size() == 0; }
     detail::tuple_accessor operator[](size_t index) const { return {*this, index}; }
-    detail::item_accessor operator[](handle h) const { return object::operator[](h); }
+    template <typename T, detail::enable_if_t<detail::is_pyobject<T>::value, int> = 0>
+    detail::item_accessor operator[](T &&o) const {
+        return object::operator[](std::forward<T>(o));
+    }
     detail::tuple_iterator begin() const { return {*this, 0}; }
     detail::tuple_iterator end() const { return {*this, PyTuple_GET_SIZE(m_ptr)}; }
 };
@@ -1692,7 +2126,11 @@ class dict : public object {
     void clear() /* py-non-const */ { PyDict_Clear(ptr()); }
     template <typename T>
     bool contains(T &&key) const {
-        return PyDict_Contains(m_ptr, detail::object_or_cast(std::forward<T>(key)).ptr()) == 1;
+        auto result = PyDict_Contains(m_ptr, detail::object_or_cast(std::forward<T>(key)).ptr());
+        if (result == -1) {
+            throw error_already_set();
+        }
+        return result == 1;
     }
 
 private:
@@ -1717,7 +2155,10 @@ class sequence : public object {
     }
     bool empty() const { return size() == 0; }
     detail::sequence_accessor operator[](size_t index) const { return {*this, index}; }
-    detail::item_accessor operator[](handle h) const { return object::operator[](h); }
+    template <typename T, detail::enable_if_t<detail::is_pyobject<T>::value, int> = 0>
+    detail::item_accessor operator[](T &&o) const {
+        return object::operator[](std::forward<T>(o));
+    }
     detail::sequence_iterator begin() const { return {*this, 0}; }
     detail::sequence_iterator end() const { return {*this, PySequence_Size(m_ptr)}; }
 };
@@ -1736,19 +2177,33 @@ class list : public object {
     size_t size() const { return (size_t) PyList_Size(m_ptr); }
     bool empty() const { return size() == 0; }
     detail::list_accessor operator[](size_t index) const { return {*this, index}; }
-    detail::item_accessor operator[](handle h) const { return object::operator[](h); }
+    template <typename T, detail::enable_if_t<detail::is_pyobject<T>::value, int> = 0>
+    detail::item_accessor operator[](T &&o) const {
+        return object::operator[](std::forward<T>(o));
+    }
     detail::list_iterator begin() const { return {*this, 0}; }
     detail::list_iterator end() const { return {*this, PyList_GET_SIZE(m_ptr)}; }
     template <typename T>
     void append(T &&val) /* py-non-const */ {
-        PyList_Append(m_ptr, detail::object_or_cast(std::forward<T>(val)).ptr());
+        if (PyList_Append(m_ptr, detail::object_or_cast(std::forward<T>(val)).ptr()) != 0) {
+            throw error_already_set();
+        }
     }
     template <typename IdxType,
               typename ValType,
               detail::enable_if_t<std::is_integral<IdxType>::value, int> = 0>
     void insert(const IdxType &index, ValType &&val) /* py-non-const */ {
-        PyList_Insert(
-            m_ptr, ssize_t_cast(index), detail::object_or_cast(std::forward<ValType>(val)).ptr());
+        if (PyList_Insert(m_ptr,
+                          ssize_t_cast(index),
+                          detail::object_or_cast(std::forward<ValType>(val)).ptr())
+            != 0) {
+            throw error_already_set();
+        }
+    }
+    void clear() /* py-non-const */ {
+        if (PyList_SetSlice(m_ptr, 0, PyList_Size(m_ptr), nullptr) == -1) {
+            throw error_already_set();
+        }
     }
 };
 
@@ -1759,25 +2214,39 @@ class kwargs : public dict {
     PYBIND11_OBJECT_DEFAULT(kwargs, dict, PyDict_Check)
 };
 
-class set : public object {
+class anyset : public object {
+public:
+    PYBIND11_OBJECT(anyset, object, PyAnySet_Check)
+    size_t size() const { return static_cast<size_t>(PySet_Size(m_ptr)); }
+    bool empty() const { return size() == 0; }
+    template <typename T>
+    bool contains(T &&val) const {
+        auto result = PySet_Contains(m_ptr, detail::object_or_cast(std::forward<T>(val)).ptr());
+        if (result == -1) {
+            throw error_already_set();
+        }
+        return result == 1;
+    }
+};
+
+class set : public anyset {
 public:
-    PYBIND11_OBJECT_CVT(set, object, PySet_Check, PySet_New)
-    set() : object(PySet_New(nullptr), stolen_t{}) {
+    PYBIND11_OBJECT_CVT(set, anyset, PySet_Check, PySet_New)
+    set() : anyset(PySet_New(nullptr), stolen_t{}) {
         if (!m_ptr) {
             pybind11_fail("Could not allocate set object!");
         }
     }
-    size_t size() const { return (size_t) PySet_Size(m_ptr); }
-    bool empty() const { return size() == 0; }
     template <typename T>
     bool add(T &&val) /* py-non-const */ {
         return PySet_Add(m_ptr, detail::object_or_cast(std::forward<T>(val)).ptr()) == 0;
     }
     void clear() /* py-non-const */ { PySet_Clear(m_ptr); }
-    template <typename T>
-    bool contains(T &&val) const {
-        return PySet_Contains(m_ptr, detail::object_or_cast(std::forward<T>(val)).ptr()) == 1;
-    }
+};
+
+class frozenset : public anyset {
+public:
+    PYBIND11_OBJECT_CVT(frozenset, anyset, PyFrozenSet_Check, PyFrozenSet_New)
 };
 
 class function : public object {
@@ -1889,8 +2358,8 @@ class memoryview : public object {
         return memoryview::from_buffer(reinterpret_cast<void *>(ptr),
                                        sizeof(T),
                                        format_descriptor<T>::value,
-                                       shape,
-                                       strides,
+                                       std::move(shape),
+                                       std::move(strides),
                                        readonly);
     }
 
@@ -1898,7 +2367,8 @@ class memoryview : public object {
     static memoryview from_buffer(const T *ptr,
                                   detail::any_container<ssize_t> shape,
                                   detail::any_container<ssize_t> strides) {
-        return memoryview::from_buffer(const_cast<T *>(ptr), shape, strides, true);
+        return memoryview::from_buffer(
+            const_cast<T *>(ptr), std::move(shape), std::move(strides), true);
     }
 
     /** \rst
@@ -2025,6 +2495,10 @@ item_accessor object_api<D>::operator[](handle key) const {
     return {derived(), reinterpret_borrow<object>(key)};
 }
 template <typename D>
+item_accessor object_api<D>::operator[](object &&key) const {
+    return {derived(), std::move(key)};
+}
+template <typename D>
 item_accessor object_api<D>::operator[](const char *key) const {
     return {derived(), pybind11::str(key)};
 }
@@ -2033,6 +2507,10 @@ obj_attr_accessor object_api<D>::attr(handle key) const {
     return {derived(), reinterpret_borrow<object>(key)};
 }
 template <typename D>
+obj_attr_accessor object_api<D>::attr(object &&key) const {
+    return {derived(), std::move(key)};
+}
+template <typename D>
 str_attr_accessor object_api<D>::attr(const char *key) const {
     return {derived(), key};
 }
@@ -2088,29 +2566,39 @@ bool object_api<D>::rich_compare(object_api const &other, int value) const {
         return result;                                                                            \
     }
 
+#define PYBIND11_MATH_OPERATOR_BINARY_INPLACE(iop, fn)                                            \
+    template <typename D>                                                                         \
+    object object_api<D>::iop(object_api const &other) {                                          \
+        object result = reinterpret_steal<object>(fn(derived().ptr(), other.derived().ptr()));    \
+        if (!result.ptr())                                                                        \
+            throw error_already_set();                                                            \
+        return result;                                                                            \
+    }
+
 PYBIND11_MATH_OPERATOR_UNARY(operator~, PyNumber_Invert)
 PYBIND11_MATH_OPERATOR_UNARY(operator-, PyNumber_Negative)
 PYBIND11_MATH_OPERATOR_BINARY(operator+, PyNumber_Add)
-PYBIND11_MATH_OPERATOR_BINARY(operator+=, PyNumber_InPlaceAdd)
+PYBIND11_MATH_OPERATOR_BINARY_INPLACE(operator+=, PyNumber_InPlaceAdd)
 PYBIND11_MATH_OPERATOR_BINARY(operator-, PyNumber_Subtract)
-PYBIND11_MATH_OPERATOR_BINARY(operator-=, PyNumber_InPlaceSubtract)
+PYBIND11_MATH_OPERATOR_BINARY_INPLACE(operator-=, PyNumber_InPlaceSubtract)
 PYBIND11_MATH_OPERATOR_BINARY(operator*, PyNumber_Multiply)
-PYBIND11_MATH_OPERATOR_BINARY(operator*=, PyNumber_InPlaceMultiply)
+PYBIND11_MATH_OPERATOR_BINARY_INPLACE(operator*=, PyNumber_InPlaceMultiply)
 PYBIND11_MATH_OPERATOR_BINARY(operator/, PyNumber_TrueDivide)
-PYBIND11_MATH_OPERATOR_BINARY(operator/=, PyNumber_InPlaceTrueDivide)
+PYBIND11_MATH_OPERATOR_BINARY_INPLACE(operator/=, PyNumber_InPlaceTrueDivide)
 PYBIND11_MATH_OPERATOR_BINARY(operator|, PyNumber_Or)
-PYBIND11_MATH_OPERATOR_BINARY(operator|=, PyNumber_InPlaceOr)
+PYBIND11_MATH_OPERATOR_BINARY_INPLACE(operator|=, PyNumber_InPlaceOr)
 PYBIND11_MATH_OPERATOR_BINARY(operator&, PyNumber_And)
-PYBIND11_MATH_OPERATOR_BINARY(operator&=, PyNumber_InPlaceAnd)
+PYBIND11_MATH_OPERATOR_BINARY_INPLACE(operator&=, PyNumber_InPlaceAnd)
 PYBIND11_MATH_OPERATOR_BINARY(operator^, PyNumber_Xor)
-PYBIND11_MATH_OPERATOR_BINARY(operator^=, PyNumber_InPlaceXor)
+PYBIND11_MATH_OPERATOR_BINARY_INPLACE(operator^=, PyNumber_InPlaceXor)
 PYBIND11_MATH_OPERATOR_BINARY(operator<<, PyNumber_Lshift)
-PYBIND11_MATH_OPERATOR_BINARY(operator<<=, PyNumber_InPlaceLshift)
+PYBIND11_MATH_OPERATOR_BINARY_INPLACE(operator<<=, PyNumber_InPlaceLshift)
 PYBIND11_MATH_OPERATOR_BINARY(operator>>, PyNumber_Rshift)
-PYBIND11_MATH_OPERATOR_BINARY(operator>>=, PyNumber_InPlaceRshift)
+PYBIND11_MATH_OPERATOR_BINARY_INPLACE(operator>>=, PyNumber_InPlaceRshift)
 
 #undef PYBIND11_MATH_OPERATOR_UNARY
 #undef PYBIND11_MATH_OPERATOR_BINARY
+#undef PYBIND11_MATH_OPERATOR_BINARY_INPLACE
 
 PYBIND11_NAMESPACE_END(detail)
 PYBIND11_NAMESPACE_END(PYBIND11_NAMESPACE)
diff --git a/include/pybind11/stl.h b/include/pybind11/stl.h
index 3d1ca7ac23..71bc5902ef 100644
--- a/include/pybind11/stl.h
+++ b/include/pybind11/stl.h
@@ -13,9 +13,9 @@
 #include "detail/common.h"
 
 #include <deque>
-#include <iostream>
 #include <list>
 #include <map>
+#include <ostream>
 #include <set>
 #include <unordered_map>
 #include <unordered_set>
@@ -45,21 +45,35 @@ using forwarded_type = conditional_t<std::is_lvalue_reference<T>::value,
 /// Forwards a value U as rvalue or lvalue according to whether T is rvalue or lvalue; typically
 /// used for forwarding a container's elements.
 template <typename T, typename U>
-forwarded_type<T, U> forward_like(U &&u) {
+constexpr forwarded_type<T, U> forward_like(U &&u) {
     return std::forward<detail::forwarded_type<T, U>>(std::forward<U>(u));
 }
 
+// Checks if a container has a STL style reserve method.
+// This will only return true for a `reserve()` with a `void` return.
+template <typename C>
+using has_reserve_method = std::is_same<decltype(std::declval<C>().reserve(0)), void>;
+
 template <typename Type, typename Key>
 struct set_caster {
     using type = Type;
     using key_conv = make_caster<Key>;
 
+private:
+    template <typename T = Type, enable_if_t<has_reserve_method<T>::value, int> = 0>
+    void reserve_maybe(const anyset &s, Type *) {
+        value.reserve(s.size());
+    }
+    void reserve_maybe(const anyset &, void *) {}
+
+public:
     bool load(handle src, bool convert) {
-        if (!isinstance<pybind11::set>(src)) {
+        if (!isinstance<anyset>(src)) {
             return false;
         }
-        auto s = reinterpret_borrow<pybind11::set>(src);
+        auto s = reinterpret_borrow<anyset>(src);
         value.clear();
+        reserve_maybe(s, &value);
         for (auto entry : s) {
             key_conv conv;
             if (!conv.load(entry, convert)) {
@@ -78,15 +92,15 @@ struct set_caster {
         pybind11::set s;
         for (auto &&value : src) {
             auto value_ = reinterpret_steal<object>(
-                key_conv::cast(forward_like<T>(value), policy, parent));
-            if (!value_ || !s.add(value_)) {
+                key_conv::cast(detail::forward_like<T>(value), policy, parent));
+            if (!value_ || !s.add(std::move(value_))) {
                 return handle();
             }
         }
         return s.release();
     }
 
-    PYBIND11_TYPE_CASTER(type, const_name("Set[") + key_conv::name + const_name("]"));
+    PYBIND11_TYPE_CASTER(type, const_name("set[") + key_conv::name + const_name("]"));
 };
 
 template <typename Type, typename Key, typename Value>
@@ -94,12 +108,21 @@ struct map_caster {
     using key_conv = make_caster<Key>;
     using value_conv = make_caster<Value>;
 
+private:
+    template <typename T = Type, enable_if_t<has_reserve_method<T>::value, int> = 0>
+    void reserve_maybe(const dict &d, Type *) {
+        value.reserve(d.size());
+    }
+    void reserve_maybe(const dict &, void *) {}
+
+public:
     bool load(handle src, bool convert) {
         if (!isinstance<dict>(src)) {
             return false;
         }
         auto d = reinterpret_borrow<dict>(src);
         value.clear();
+        reserve_maybe(d, &value);
         for (auto it : d) {
             key_conv kconv;
             value_conv vconv;
@@ -122,19 +145,19 @@ struct map_caster {
         }
         for (auto &&kv : src) {
             auto key = reinterpret_steal<object>(
-                key_conv::cast(forward_like<T>(kv.first), policy_key, parent));
+                key_conv::cast(detail::forward_like<T>(kv.first), policy_key, parent));
             auto value = reinterpret_steal<object>(
-                value_conv::cast(forward_like<T>(kv.second), policy_value, parent));
+                value_conv::cast(detail::forward_like<T>(kv.second), policy_value, parent));
             if (!key || !value) {
                 return handle();
             }
-            d[key] = value;
+            d[std::move(key)] = std::move(value);
         }
         return d.release();
     }
 
     PYBIND11_TYPE_CASTER(Type,
-                         const_name("Dict[") + key_conv::name + const_name(", ") + value_conv::name
+                         const_name("dict[") + key_conv::name + const_name(", ") + value_conv::name
                              + const_name("]"));
 };
 
@@ -149,7 +172,7 @@ struct list_caster {
         auto s = reinterpret_borrow<sequence>(src);
         value.clear();
         reserve_maybe(s, &value);
-        for (auto it : s) {
+        for (const auto &it : s) {
             value_conv conv;
             if (!conv.load(it, convert)) {
                 return false;
@@ -160,9 +183,7 @@ struct list_caster {
     }
 
 private:
-    template <
-        typename T = Type,
-        enable_if_t<std::is_same<decltype(std::declval<T>().reserve(0)), void>::value, int> = 0>
+    template <typename T = Type, enable_if_t<has_reserve_method<T>::value, int> = 0>
     void reserve_maybe(const sequence &s, Type *) {
         value.reserve(s.size());
     }
@@ -178,7 +199,7 @@ struct list_caster {
         ssize_t index = 0;
         for (auto &&value : src) {
             auto value_ = reinterpret_steal<object>(
-                value_conv::cast(forward_like<T>(value), policy, parent));
+                value_conv::cast(detail::forward_like<T>(value), policy, parent));
             if (!value_) {
                 return handle();
             }
@@ -187,7 +208,7 @@ struct list_caster {
         return l.release();
     }
 
-    PYBIND11_TYPE_CASTER(Type, const_name("List[") + value_conv::name + const_name("]"));
+    PYBIND11_TYPE_CASTER(Type, const_name("list[") + value_conv::name + const_name("]"));
 };
 
 template <typename Type, typename Alloc>
@@ -226,7 +247,7 @@ struct array_caster {
             return false;
         }
         size_t ctr = 0;
-        for (auto it : l) {
+        for (const auto &it : l) {
             value_conv conv;
             if (!conv.load(it, convert)) {
                 return false;
@@ -242,7 +263,7 @@ struct array_caster {
         ssize_t index = 0;
         for (auto &&value : src) {
             auto value_ = reinterpret_steal<object>(
-                value_conv::cast(forward_like<T>(value), policy, parent));
+                value_conv::cast(detail::forward_like<T>(value), policy, parent));
             if (!value_) {
                 return handle();
             }
@@ -252,11 +273,11 @@ struct array_caster {
     }
 
     PYBIND11_TYPE_CASTER(ArrayType,
-                         const_name("List[") + value_conv::name
+                         const_name<Resizable>(const_name(""), const_name("Annotated["))
+                             + const_name("list[") + value_conv::name + const_name("]")
                              + const_name<Resizable>(const_name(""),
-                                                     const_name("[") + const_name<Size>()
-                                                         + const_name("]"))
-                             + const_name("]"));
+                                                     const_name(", FixedSize(")
+                                                         + const_name<Size>() + const_name(")]")));
 };
 
 template <typename Type, size_t Size>
@@ -290,11 +311,12 @@ struct optional_caster {
     template <typename T>
     static handle cast(T &&src, return_value_policy policy, handle parent) {
         if (!src) {
-            return none().inc_ref();
+            return none().release();
         }
         if (!std::is_lvalue_reference<T>::value) {
             policy = return_value_policy_override<Value>::policy(policy);
         }
+        // NOLINTNEXTLINE(bugprone-unchecked-optional-access)
         return value_conv::cast(*std::forward<T>(src), policy, parent);
     }
 
@@ -399,7 +421,8 @@ struct variant_caster<V<Ts...>> {
 
     using Type = V<Ts...>;
     PYBIND11_TYPE_CASTER(Type,
-                         const_name("Union[") + detail::concat(make_caster<Ts>::name...)
+                         const_name("Union[")
+                             + ::pybind11::detail::concat(make_caster<Ts>::name...)
                              + const_name("]"));
 };
 
diff --git a/include/pybind11/stl/filesystem.h b/include/pybind11/stl/filesystem.h
index 93727424c9..85c131efe4 100644
--- a/include/pybind11/stl/filesystem.h
+++ b/include/pybind11/stl/filesystem.h
@@ -13,21 +13,27 @@
 #include <string>
 
 #ifdef __has_include
-#    if defined(PYBIND11_CPP17) && __has_include(<filesystem>)
-#        include <filesystem>
-#        define PYBIND11_HAS_FILESYSTEM 1
+#    if defined(PYBIND11_CPP17)
+#        if __has_include(<filesystem>)
+#            include <filesystem>
+#            define PYBIND11_HAS_FILESYSTEM 1
+#        elif __has_include(<experimental/filesystem>)
+#            include <experimental/filesystem>
+#            define PYBIND11_HAS_EXPERIMENTAL_FILESYSTEM 1
+#        endif
 #    endif
 #endif
 
-#if !defined(PYBIND11_HAS_FILESYSTEM) && !defined(PYBIND11_HAS_FILESYSTEM_IS_OPTIONAL)
+#if !defined(PYBIND11_HAS_FILESYSTEM) && !defined(PYBIND11_HAS_EXPERIMENTAL_FILESYSTEM)           \
+    && !defined(PYBIND11_HAS_FILESYSTEM_IS_OPTIONAL)
 #    error                                                                                        \
-        "#include <filesystem> is not available. (Use -DPYBIND11_HAS_FILESYSTEM_IS_OPTIONAL to ignore.)"
+        "Neither #include <filesystem> nor #include <experimental/filesystem is available. (Use -DPYBIND11_HAS_FILESYSTEM_IS_OPTIONAL to ignore.)"
 #endif
 
 PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)
 PYBIND11_NAMESPACE_BEGIN(detail)
 
-#if defined(PYBIND11_HAS_FILESYSTEM)
+#if defined(PYBIND11_HAS_FILESYSTEM) || defined(PYBIND11_HAS_EXPERIMENTAL_FILESYSTEM)
 template <typename T>
 struct path_caster {
 
@@ -94,9 +100,16 @@ struct path_caster {
     PYBIND11_TYPE_CASTER(T, const_name("os.PathLike"));
 };
 
+#endif // PYBIND11_HAS_FILESYSTEM || defined(PYBIND11_HAS_EXPERIMENTAL_FILESYSTEM)
+
+#if defined(PYBIND11_HAS_FILESYSTEM)
 template <>
 struct type_caster<std::filesystem::path> : public path_caster<std::filesystem::path> {};
-#endif // PYBIND11_HAS_FILESYSTEM
+#elif defined(PYBIND11_HAS_EXPERIMENTAL_FILESYSTEM)
+template <>
+struct type_caster<std::experimental::filesystem::path>
+    : public path_caster<std::experimental::filesystem::path> {};
+#endif
 
 PYBIND11_NAMESPACE_END(detail)
 PYBIND11_NAMESPACE_END(PYBIND11_NAMESPACE)
diff --git a/include/pybind11/stl_bind.h b/include/pybind11/stl_bind.h
index 3a3e54dd48..66c452ea79 100644
--- a/include/pybind11/stl_bind.h
+++ b/include/pybind11/stl_bind.h
@@ -10,10 +10,13 @@
 #pragma once
 
 #include "detail/common.h"
+#include "detail/type_caster_base.h"
+#include "cast.h"
 #include "operators.h"
 
 #include <algorithm>
 #include <sstream>
+#include <type_traits>
 
 PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)
 PYBIND11_NAMESPACE_BEGIN(detail)
@@ -58,9 +61,11 @@ struct is_comparable<
 /* For a vector/map data structure, recursively check the value type
    (which is std::pair for maps) */
 template <typename T>
-struct is_comparable<T, enable_if_t<container_traits<T>::is_vector>> {
-    static constexpr const bool value = is_comparable<typename T::value_type>::value;
-};
+struct is_comparable<T, enable_if_t<container_traits<T>::is_vector>>
+    : is_comparable<typename recursive_container_traits<T>::type_to_check_recursively> {};
+
+template <>
+struct is_comparable<recursive_bottom> : std::true_type {};
 
 /* For pairs, recursively check the two data types */
 template <typename T>
@@ -153,8 +158,7 @@ void vector_modifiers(
         return v.release();
     }));
 
-    cl.def(
-        "clear", [](Vector &v) { v.clear(); }, "Clear the contents");
+    cl.def("clear", [](Vector &v) { v.clear(); }, "Clear the contents");
 
     cl.def(
         "extend",
@@ -232,7 +236,7 @@ void vector_modifiers(
     /// Slicing protocol
     cl.def(
         "__getitem__",
-        [](const Vector &v, slice slice) -> Vector * {
+        [](const Vector &v, const slice &slice) -> Vector * {
             size_t start = 0, stop = 0, step = 0, slicelength = 0;
 
             if (!slice.compute(v.size(), &start, &stop, &step, &slicelength)) {
@@ -253,7 +257,7 @@ void vector_modifiers(
 
     cl.def(
         "__setitem__",
-        [](Vector &v, slice slice, const Vector &value) {
+        [](Vector &v, const slice &slice, const Vector &value) {
             size_t start = 0, stop = 0, step = 0, slicelength = 0;
             if (!slice.compute(v.size(), &start, &stop, &step, &slicelength)) {
                 throw error_already_set();
@@ -281,7 +285,7 @@ void vector_modifiers(
 
     cl.def(
         "__delitem__",
-        [](Vector &v, slice slice) {
+        [](Vector &v, const slice &slice) {
             size_t start = 0, stop = 0, step = 0, slicelength = 0;
 
             if (!slice.compute(v.size(), &start, &stop, &step, &slicelength)) {
@@ -352,13 +356,17 @@ void vector_accessor(enable_if_t<vector_needs_copy<Vector>::value, Class_> &cl)
     using DiffType = typename Vector::difference_type;
     using ItType = typename Vector::iterator;
     cl.def("__getitem__", [](const Vector &v, DiffType i) -> T {
-        if (i < 0 && (i += v.size()) < 0) {
-            throw index_error();
+        if (i < 0) {
+            i += v.size();
+            if (i < 0) {
+                throw index_error();
+            }
         }
-        if ((SizeType) i >= v.size()) {
+        auto i_st = static_cast<SizeType>(i);
+        if (i_st >= v.size()) {
             throw index_error();
         }
-        return v[(SizeType) i];
+        return v[i_st];
     });
 
     cl.def(
@@ -516,7 +524,7 @@ class_<Vector, holder_type> bind_vector(handle scope, std::string const &name, A
         [](const Vector &v) -> bool { return !v.empty(); },
         "Check whether the list is nonempty");
 
-    cl.def("__len__", &Vector::size);
+    cl.def("__len__", [](const Vector &vec) { return vec.size(); });
 
 #if 0
     // C++ style functions deprecated, leaving it here as an example
@@ -636,18 +644,53 @@ auto map_if_insertion_operator(Class_ &cl, std::string const &name)
         "Return the canonical string representation of this map.");
 }
 
-template <typename Map>
 struct keys_view {
+    virtual size_t len() = 0;
+    virtual iterator iter() = 0;
+    virtual bool contains(const handle &k) = 0;
+    virtual ~keys_view() = default;
+};
+
+struct values_view {
+    virtual size_t len() = 0;
+    virtual iterator iter() = 0;
+    virtual ~values_view() = default;
+};
+
+struct items_view {
+    virtual size_t len() = 0;
+    virtual iterator iter() = 0;
+    virtual ~items_view() = default;
+};
+
+template <typename Map>
+struct KeysViewImpl : public detail::keys_view {
+    explicit KeysViewImpl(Map &map) : map(map) {}
+    size_t len() override { return map.size(); }
+    iterator iter() override { return make_key_iterator(map.begin(), map.end()); }
+    bool contains(const handle &k) override {
+        try {
+            return map.find(k.template cast<typename Map::key_type>()) != map.end();
+        } catch (const cast_error &) {
+            return false;
+        }
+    }
     Map &map;
 };
 
 template <typename Map>
-struct values_view {
+struct ValuesViewImpl : public detail::values_view {
+    explicit ValuesViewImpl(Map &map) : map(map) {}
+    size_t len() override { return map.size(); }
+    iterator iter() override { return make_value_iterator(map.begin(), map.end()); }
     Map &map;
 };
 
 template <typename Map>
-struct items_view {
+struct ItemsViewImpl : public detail::items_view {
+    explicit ItemsViewImpl(Map &map) : map(map) {}
+    size_t len() override { return map.size(); }
+    iterator iter() override { return make_iterator(map.begin(), map.end()); }
     Map &map;
 };
 
@@ -657,9 +700,9 @@ template <typename Map, typename holder_type = std::unique_ptr<Map>, typename...
 class_<Map, holder_type> bind_map(handle scope, const std::string &name, Args &&...args) {
     using KeyType = typename Map::key_type;
     using MappedType = typename Map::mapped_type;
-    using KeysView = detail::keys_view<Map>;
-    using ValuesView = detail::values_view<Map>;
-    using ItemsView = detail::items_view<Map>;
+    using KeysView = detail::keys_view;
+    using ValuesView = detail::values_view;
+    using ItemsView = detail::items_view;
     using Class_ = class_<Map, holder_type>;
 
     // If either type is a non-module-local bound type then make the map binding non-local as well;
@@ -673,12 +716,35 @@ class_<Map, holder_type> bind_map(handle scope, const std::string &name, Args &&
     }
 
     Class_ cl(scope, name.c_str(), pybind11::module_local(local), std::forward<Args>(args)...);
-    class_<KeysView> keys_view(
-        scope, ("KeysView[" + name + "]").c_str(), pybind11::module_local(local));
-    class_<ValuesView> values_view(
-        scope, ("ValuesView[" + name + "]").c_str(), pybind11::module_local(local));
-    class_<ItemsView> items_view(
-        scope, ("ItemsView[" + name + "]").c_str(), pybind11::module_local(local));
+
+    // Wrap KeysView if it wasn't already wrapped
+    if (!detail::get_type_info(typeid(KeysView))) {
+        class_<KeysView> keys_view(scope, "KeysView", pybind11::module_local(local));
+        keys_view.def("__len__", &KeysView::len);
+        keys_view.def("__iter__",
+                      &KeysView::iter,
+                      keep_alive<0, 1>() /* Essential: keep view alive while iterator exists */
+        );
+        keys_view.def("__contains__", &KeysView::contains);
+    }
+    // Similarly for ValuesView:
+    if (!detail::get_type_info(typeid(ValuesView))) {
+        class_<ValuesView> values_view(scope, "ValuesView", pybind11::module_local(local));
+        values_view.def("__len__", &ValuesView::len);
+        values_view.def("__iter__",
+                        &ValuesView::iter,
+                        keep_alive<0, 1>() /* Essential: keep view alive while iterator exists */
+        );
+    }
+    // Similarly for ItemsView:
+    if (!detail::get_type_info(typeid(ItemsView))) {
+        class_<ItemsView> items_view(scope, "ItemsView", pybind11::module_local(local));
+        items_view.def("__len__", &ItemsView::len);
+        items_view.def("__iter__",
+                       &ItemsView::iter,
+                       keep_alive<0, 1>() /* Essential: keep view alive while iterator exists */
+        );
+    }
 
     cl.def(init<>());
 
@@ -698,19 +764,19 @@ class_<Map, holder_type> bind_map(handle scope, const std::string &name, Args &&
 
     cl.def(
         "keys",
-        [](Map &m) { return KeysView{m}; },
+        [](Map &m) { return std::unique_ptr<KeysView>(new detail::KeysViewImpl<Map>(m)); },
         keep_alive<0, 1>() /* Essential: keep map alive while view exists */
     );
 
     cl.def(
         "values",
-        [](Map &m) { return ValuesView{m}; },
+        [](Map &m) { return std::unique_ptr<ValuesView>(new detail::ValuesViewImpl<Map>(m)); },
         keep_alive<0, 1>() /* Essential: keep map alive while view exists */
     );
 
     cl.def(
         "items",
-        [](Map &m) { return ItemsView{m}; },
+        [](Map &m) { return std::unique_ptr<ItemsView>(new detail::ItemsViewImpl<Map>(m)); },
         keep_alive<0, 1>() /* Essential: keep map alive while view exists */
     );
 
@@ -747,37 +813,8 @@ class_<Map, holder_type> bind_map(handle scope, const std::string &name, Args &&
         m.erase(it);
     });
 
-    cl.def("__len__", &Map::size);
-
-    keys_view.def("__len__", [](KeysView &view) { return view.map.size(); });
-    keys_view.def(
-        "__iter__",
-        [](KeysView &view) { return make_key_iterator(view.map.begin(), view.map.end()); },
-        keep_alive<0, 1>() /* Essential: keep view alive while iterator exists */
-    );
-    keys_view.def("__contains__", [](KeysView &view, const KeyType &k) -> bool {
-        auto it = view.map.find(k);
-        if (it == view.map.end()) {
-            return false;
-        }
-        return true;
-    });
-    // Fallback for when the object is not of the key type
-    keys_view.def("__contains__", [](KeysView &, const object &) -> bool { return false; });
-
-    values_view.def("__len__", [](ValuesView &view) { return view.map.size(); });
-    values_view.def(
-        "__iter__",
-        [](ValuesView &view) { return make_value_iterator(view.map.begin(), view.map.end()); },
-        keep_alive<0, 1>() /* Essential: keep view alive while iterator exists */
-    );
-
-    items_view.def("__len__", [](ItemsView &view) { return view.map.size(); });
-    items_view.def(
-        "__iter__",
-        [](ItemsView &view) { return make_iterator(view.map.begin(), view.map.end()); },
-        keep_alive<0, 1>() /* Essential: keep view alive while iterator exists */
-    );
+    // Always use a lambda in case of `using` declaration
+    cl.def("__len__", [](const Map &m) { return m.size(); });
 
     return cl;
 }
diff --git a/include/pybind11/type_caster_pyobject_ptr.h b/include/pybind11/type_caster_pyobject_ptr.h
new file mode 100644
index 0000000000..aa914f9e15
--- /dev/null
+++ b/include/pybind11/type_caster_pyobject_ptr.h
@@ -0,0 +1,61 @@
+// Copyright (c) 2023 The pybind Community.
+
+#pragma once
+
+#include "detail/common.h"
+#include "detail/descr.h"
+#include "cast.h"
+#include "pytypes.h"
+
+PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)
+PYBIND11_NAMESPACE_BEGIN(detail)
+
+template <>
+class type_caster<PyObject> {
+public:
+    static constexpr auto name = const_name("object"); // See discussion under PR #4601.
+
+    // This overload is purely to guard against accidents.
+    template <typename T,
+              detail::enable_if_t<!is_same_ignoring_cvref<T, PyObject *>::value, int> = 0>
+    static handle cast(T &&, return_value_policy, handle /*parent*/) {
+        static_assert(is_same_ignoring_cvref<T, PyObject *>::value,
+                      "Invalid C++ type T for to-Python conversion (type_caster<PyObject>).");
+        return nullptr; // Unreachable.
+    }
+
+    static handle cast(PyObject *src, return_value_policy policy, handle /*parent*/) {
+        if (src == nullptr) {
+            throw error_already_set();
+        }
+        if (PyErr_Occurred()) {
+            raise_from(PyExc_SystemError, "src != nullptr but PyErr_Occurred()");
+            throw error_already_set();
+        }
+        if (policy == return_value_policy::take_ownership) {
+            return src;
+        }
+        if (policy == return_value_policy::reference
+            || policy == return_value_policy::automatic_reference) {
+            return handle(src).inc_ref();
+        }
+        pybind11_fail("type_caster<PyObject>::cast(): unsupported return_value_policy: "
+                      + std::to_string(static_cast<int>(policy)));
+    }
+
+    bool load(handle src, bool) {
+        value = reinterpret_borrow<object>(src);
+        return true;
+    }
+
+    template <typename T>
+    using cast_op_type = PyObject *;
+
+    explicit operator PyObject *() { return value.ptr(); }
+
+private:
+    object value;
+};
+
+PYBIND11_NAMESPACE_END(detail)
+PYBIND11_NAMESPACE_END(PYBIND11_NAMESPACE)
diff --git a/include/pybind11/typing.h b/include/pybind11/typing.h
new file mode 100644
index 0000000000..b0feb9464a
--- /dev/null
+++ b/include/pybind11/typing.h
@@ -0,0 +1,244 @@
+/*
+    pybind11/typing.h: Convenience wrapper classes for basic Python types
+    with more explicit annotations.
+
+    Copyright (c) 2023 Dustin Spicuzza <dustin@virtualroadside.com>
+
+    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 "detail/common.h"
+#include "cast.h"
+#include "pytypes.h"
+
+#include <algorithm>
+
+PYBIND11_NAMESPACE_BEGIN(PYBIND11_NAMESPACE)
+PYBIND11_NAMESPACE_BEGIN(typing)
+
+/*
+    The following types can be used to direct pybind11-generated docstrings
+    to have have more explicit types (e.g., `list[str]` instead of `list`).
+    Just use these in place of existing types.
+
+    There is no additional enforcement of types at runtime.
+*/
+
+template <typename... Types>
+class Tuple : public tuple {
+    using tuple::tuple;
+};
+
+template <typename K, typename V>
+class Dict : public dict {
+    using dict::dict;
+};
+
+template <typename T>
+class List : public list {
+    using list::list;
+};
+
+template <typename T>
+class Set : public set {
+    using set::set;
+};
+
+template <typename T>
+class Iterable : public iterable {
+    using iterable::iterable;
+};
+
+template <typename T>
+class Iterator : public iterator {
+    using iterator::iterator;
+};
+
+template <typename Signature>
+class Callable;
+
+template <typename Return, typename... Args>
+class Callable<Return(Args...)> : public function {
+    using function::function;
+};
+
+template <typename T>
+class Type : public type {
+    using type::type;
+};
+
+template <typename... Types>
+class Union : public object {
+    PYBIND11_OBJECT_DEFAULT(Union, object, PyObject_Type)
+    using object::object;
+};
+
+template <typename T>
+class Optional : public object {
+    PYBIND11_OBJECT_DEFAULT(Optional, object, PyObject_Type)
+    using object::object;
+};
+
+template <typename T>
+class TypeGuard : public bool_ {
+    using bool_::bool_;
+};
+
+template <typename T>
+class TypeIs : public bool_ {
+    using bool_::bool_;
+};
+
+class NoReturn : public none {
+    using none::none;
+};
+
+class Never : public none {
+    using none::none;
+};
+
+#if defined(__cpp_nontype_template_parameter_class)                                               \
+    && (/* See #5201 */ !defined(__GNUC__)                                                        \
+        || (__GNUC__ > 10 || (__GNUC__ == 10 && __GNUC_MINOR__ >= 3)))
+#    define PYBIND11_TYPING_H_HAS_STRING_LITERAL
+template <size_t N>
+struct StringLiteral {
+    constexpr StringLiteral(const char (&str)[N]) { std::copy_n(str, N, name); }
+    char name[N];
+};
+
+template <StringLiteral... StrLits>
+class Literal : public object {
+    PYBIND11_OBJECT_DEFAULT(Literal, object, PyObject_Type)
+};
+
+// Example syntax for creating a TypeVar.
+// typedef typing::TypeVar<"T"> TypeVarT;
+template <StringLiteral>
+class TypeVar : public object {
+    PYBIND11_OBJECT_DEFAULT(TypeVar, object, PyObject_Type)
+    using object::object;
+};
+#endif
+
+PYBIND11_NAMESPACE_END(typing)
+
+PYBIND11_NAMESPACE_BEGIN(detail)
+
+template <typename... Types>
+struct handle_type_name<typing::Tuple<Types...>> {
+    static constexpr auto name = const_name("tuple[")
+                                 + ::pybind11::detail::concat(make_caster<Types>::name...)
+                                 + const_name("]");
+};
+
+template <>
+struct handle_type_name<typing::Tuple<>> {
+    // PEP 484 specifies this syntax for an empty tuple
+    static constexpr auto name = const_name("tuple[()]");
+};
+
+template <typename T>
+struct handle_type_name<typing::Tuple<T, ellipsis>> {
+    // PEP 484 specifies this syntax for a variable-length tuple
+    static constexpr auto name
+        = const_name("tuple[") + make_caster<T>::name + const_name(", ...]");
+};
+
+template <typename K, typename V>
+struct handle_type_name<typing::Dict<K, V>> {
+    static constexpr auto name = const_name("dict[") + make_caster<K>::name + const_name(", ")
+                                 + make_caster<V>::name + const_name("]");
+};
+
+template <typename T>
+struct handle_type_name<typing::List<T>> {
+    static constexpr auto name = const_name("list[") + make_caster<T>::name + const_name("]");
+};
+
+template <typename T>
+struct handle_type_name<typing::Set<T>> {
+    static constexpr auto name = const_name("set[") + make_caster<T>::name + const_name("]");
+};
+
+template <typename T>
+struct handle_type_name<typing::Iterable<T>> {
+    static constexpr auto name = const_name("Iterable[") + make_caster<T>::name + const_name("]");
+};
+
+template <typename T>
+struct handle_type_name<typing::Iterator<T>> {
+    static constexpr auto name = const_name("Iterator[") + make_caster<T>::name + const_name("]");
+};
+
+template <typename Return, typename... Args>
+struct handle_type_name<typing::Callable<Return(Args...)>> {
+    using retval_type = conditional_t<std::is_same<Return, void>::value, void_type, Return>;
+    static constexpr auto name
+        = const_name("Callable[[") + ::pybind11::detail::concat(make_caster<Args>::name...)
+          + const_name("], ") + make_caster<retval_type>::name + const_name("]");
+};
+
+template <typename Return>
+struct handle_type_name<typing::Callable<Return(ellipsis)>> {
+    // PEP 484 specifies this syntax for defining only return types of callables
+    using retval_type = conditional_t<std::is_same<Return, void>::value, void_type, Return>;
+    static constexpr auto name
+        = const_name("Callable[..., ") + make_caster<retval_type>::name + const_name("]");
+};
+
+template <typename T>
+struct handle_type_name<typing::Type<T>> {
+    static constexpr auto name = const_name("type[") + make_caster<T>::name + const_name("]");
+};
+
+template <typename... Types>
+struct handle_type_name<typing::Union<Types...>> {
+    static constexpr auto name = const_name("Union[")
+                                 + ::pybind11::detail::concat(make_caster<Types>::name...)
+                                 + const_name("]");
+};
+
+template <typename T>
+struct handle_type_name<typing::Optional<T>> {
+    static constexpr auto name = const_name("Optional[") + make_caster<T>::name + const_name("]");
+};
+
+template <typename T>
+struct handle_type_name<typing::TypeGuard<T>> {
+    static constexpr auto name = const_name("TypeGuard[") + make_caster<T>::name + const_name("]");
+};
+
+template <typename T>
+struct handle_type_name<typing::TypeIs<T>> {
+    static constexpr auto name = const_name("TypeIs[") + make_caster<T>::name + const_name("]");
+};
+
+template <>
+struct handle_type_name<typing::NoReturn> {
+    static constexpr auto name = const_name("NoReturn");
+};
+
+template <>
+struct handle_type_name<typing::Never> {
+    static constexpr auto name = const_name("Never");
+};
+
+#if defined(PYBIND11_TYPING_H_HAS_STRING_LITERAL)
+template <typing::StringLiteral... Literals>
+struct handle_type_name<typing::Literal<Literals...>> {
+    static constexpr auto name = const_name("Literal[")
+                                 + pybind11::detail::concat(const_name(Literals.name)...)
+                                 + const_name("]");
+};
+template <typing::StringLiteral StrLit>
+struct handle_type_name<typing::TypeVar<StrLit>> {
+    static constexpr auto name = const_name(StrLit.name);
+};
+#endif
+
+PYBIND11_NAMESPACE_END(detail)
+PYBIND11_NAMESPACE_END(PYBIND11_NAMESPACE)