pybind11/tests/test_class.cpp

642 lines
24 KiB
C++
Raw Normal View History

/*
tests/test_class.cpp -- test py::class_ definitions and basic functionality
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.
*/
fix: Intel ICC C++17 compatibility (#2729) * CI: Intel icc/icpc via oneAPI Add testing for Intel icc/icpc via the oneAPI images. Intel oneAPI is in a late beta stage, currently shipping oneAPI beta09 with ICC 20.2. CI: Skip Interpreter Tests for Intel Cannot find how to add this, neiter the package `libc6-dev` nor `intel-oneapi-mkl-devel` help when installed to solve this: ``` -- Looking for C++ include pthread.h -- Looking for C++ include pthread.h - not found CMake Error at /__t/cmake/3.18.4/x64/cmake-3.18.4-Linux-x86_64/share/cmake-3.18/Modules/FindPackageHandleStandardArgs.cmake:165 (message): Could NOT find Threads (missing: Threads_FOUND) Call Stack (most recent call first): /__t/cmake/3.18.4/x64/cmake-3.18.4-Linux-x86_64/share/cmake-3.18/Modules/FindPackageHandleStandardArgs.cmake:458 (_FPHSA_FAILURE_MESSAGE) /__t/cmake/3.18.4/x64/cmake-3.18.4-Linux-x86_64/share/cmake-3.18/Modules/FindThreads.cmake:234 (FIND_PACKAGE_HANDLE_STANDARD_ARGS) tests/test_embed/CMakeLists.txt:17 (find_package) ``` CI: libc6-dev from GCC for ICC CI: Run bare metal for oneAPI CI: Ubuntu 18.04 for oneAPI CI: Intel +Catch -Eigen CI: CMake from Apt (ICC tests) CI: Replace Intel Py with GCC Py CI: Intel w/o GCC's Eigen CI: ICC with verbose make [Debug] Find core dump tests: use arg{} instead of arg() for Intel tests: adding a few more missing {} fix: sync with @tobiasleibner's branch fix: try ubuntu 20-04 fix: drop exit 1 docs: Apply suggestions from code review Co-authored-by: Tobias Leibner <tobias.leibner@googlemail.com> Workaround for ICC enable_if issues Another workaround for ICC's enable_if issues fix error in previous commit Disable one test for the Intel compiler in C++17 mode Add back one instance of py::arg().noconvert() Add NOLINT to fix clang-tidy check Work around for ICC internal error in PYBIND11_EXPAND_SIDE_EFFECTS in C++17 mode CI: Intel ICC with C++17 docs: pybind11/numpy.h does not require numpy at build time. (#2720) This is nice enough to be mentioned explicitly in the docs. docs: Update warning about Python 3.9.0 UB, now that 3.9.1 has been released (#2719) Adjusting `type_caster<std::reference_wrapper<T>>` to support const/non-const propagation in `cast_op`. (#2705) * Allow type_caster of std::reference_wrapper<T> to be the same as a native reference. Before, both std::reference_wrapper<T> and std::reference_wrapper<const T> would invoke cast_op<type>. This doesn't allow the type_caster<> specialization for T to distinguish reference_wrapper types from value types. After, the type_caster<> specialization invokes cast_op<type&>, which allows reference_wrapper to behave in the same way as a native reference type. * Add tests/examples for std::reference_wrapper<const T> * Add tests which use mutable/immutable variants This test is a chimera; it blends the pybind11 casters with a custom pytype implementation that supports immutable and mutable calls. In order to detect the immutable/mutable state, the cast_op needs to propagate it, even through e.g. std::reference<const T> Note: This is still a work in progress; some things are crashing, which likely means that I have a refcounting bug or something else missing. * Add/finish tests that distinguish const& from & Fixes the bugs in my custom python type implementation, demonstrate test that requires const& and reference_wrapper<const T> being treated differently from Non-const. * Add passing a const to non-const method. * Demonstrate non-const conversion of reference_wrapper in tests. Apply formatting presubmit check. * Fix build errors from presubmit checks. * Try and fix a few more CI errors * More CI fixes. * More CI fixups. * Try and get PyPy to work. * Additional minor fixups. Getting close to CI green. * More ci fixes? * fix clang-tidy warnings from presubmit * fix more clang-tidy warnings * minor comment and consistency cleanups * PyDECREF -> Py_DECREF * copy/move constructors * Resolve codereview comments * more review comment fixes * review comments: remove spurious & * Make the test fail even when the static_assert is commented out. This expands the test_freezable_type_caster a bit by: 1/ adding accessors .is_immutable and .addr to compare identity from python. 2/ Changing the default cast_op of the type_caster<> specialization to return a non-const value. In normal codepaths this is a reasonable default. 3/ adding roundtrip variants to exercise the by reference, by pointer and by reference_wrapper in all call paths. In conjunction with 2/, this demonstrates the failure case of the existing std::reference_wrpper conversion, which now loses const in a similar way that happens when using the default cast_op_type<>. * apply presubmit formatting * Revert inclusion of test_freezable_type_caster There's some concern that this test is a bit unwieldly because of the use of the raw <Python.h> functions. Removing for now. * Add a test that validates const references propagation. This test verifies that cast_op may be used to correctly detect const reference types when used with std::reference_wrapper. * mend * Review comments based changes. 1. std::add_lvalue_reference<type> -> type& 2. Simplify the test a little more; we're never returning the ConstRefCaster type so the class_ definition can be removed. * formatted files again. * Move const_ref_caster test to builtin_casters * Review comments: use cast_op and adjust some comments. * Simplify ConstRefCasted test I like this version better as it moves the assertion that matters back into python. ci: drop pypy2 linux, PGI 20.7, add Python 10 dev (#2724) * ci: drop pypy2 linux, add Python 10 dev * ci: fix mistake * ci: commented-out PGI 20.11, drop 20.7 fix: regression with installed pybind11 overriding local one (#2716) * fix: regression with installed pybind11 overriding discovered one Closes #2709 * docs: wording incorrect style: remove redundant instance->owned = true (#2723) which was just before set to True in instance->allocate_layout() fix: also throw in the move-constructor added by the PYBIND11_OBJECT macro, after the argument has been moved-out (if necessary) (#2701) Make args_are_all_* ICC workarounds unconditional Disable test_aligned on Intel ICC Fix test_aligned on Intel ICC Skip test_python_alreadyset_in_destructor on Intel ICC Fix test_aligned again ICC CI: Downgrade pytest pytest 6 does not capture the `discard_as_unraisable` stderr and just writes a warning with its content instead. * refactor: simpler Intel workaround, suggested by @laramiel * fix: try version with impl to see if it is easier to compile * docs: update README for ICC Co-authored-by: Axel Huebl <axel.huebl@plasma.ninja> Co-authored-by: Henry Schreiner <henryschreineriii@gmail.com>
2021-01-18 00:53:07 +00:00
#if defined(__INTEL_COMPILER) && __cplusplus >= 201703L
// Intel compiler requires a separate header file to support aligned new operators
// and does not set the __cpp_aligned_new feature macro.
// This header needs to be included before pybind11.
# include <aligned_new>
fix: Intel ICC C++17 compatibility (#2729) * CI: Intel icc/icpc via oneAPI Add testing for Intel icc/icpc via the oneAPI images. Intel oneAPI is in a late beta stage, currently shipping oneAPI beta09 with ICC 20.2. CI: Skip Interpreter Tests for Intel Cannot find how to add this, neiter the package `libc6-dev` nor `intel-oneapi-mkl-devel` help when installed to solve this: ``` -- Looking for C++ include pthread.h -- Looking for C++ include pthread.h - not found CMake Error at /__t/cmake/3.18.4/x64/cmake-3.18.4-Linux-x86_64/share/cmake-3.18/Modules/FindPackageHandleStandardArgs.cmake:165 (message): Could NOT find Threads (missing: Threads_FOUND) Call Stack (most recent call first): /__t/cmake/3.18.4/x64/cmake-3.18.4-Linux-x86_64/share/cmake-3.18/Modules/FindPackageHandleStandardArgs.cmake:458 (_FPHSA_FAILURE_MESSAGE) /__t/cmake/3.18.4/x64/cmake-3.18.4-Linux-x86_64/share/cmake-3.18/Modules/FindThreads.cmake:234 (FIND_PACKAGE_HANDLE_STANDARD_ARGS) tests/test_embed/CMakeLists.txt:17 (find_package) ``` CI: libc6-dev from GCC for ICC CI: Run bare metal for oneAPI CI: Ubuntu 18.04 for oneAPI CI: Intel +Catch -Eigen CI: CMake from Apt (ICC tests) CI: Replace Intel Py with GCC Py CI: Intel w/o GCC's Eigen CI: ICC with verbose make [Debug] Find core dump tests: use arg{} instead of arg() for Intel tests: adding a few more missing {} fix: sync with @tobiasleibner's branch fix: try ubuntu 20-04 fix: drop exit 1 docs: Apply suggestions from code review Co-authored-by: Tobias Leibner <tobias.leibner@googlemail.com> Workaround for ICC enable_if issues Another workaround for ICC's enable_if issues fix error in previous commit Disable one test for the Intel compiler in C++17 mode Add back one instance of py::arg().noconvert() Add NOLINT to fix clang-tidy check Work around for ICC internal error in PYBIND11_EXPAND_SIDE_EFFECTS in C++17 mode CI: Intel ICC with C++17 docs: pybind11/numpy.h does not require numpy at build time. (#2720) This is nice enough to be mentioned explicitly in the docs. docs: Update warning about Python 3.9.0 UB, now that 3.9.1 has been released (#2719) Adjusting `type_caster<std::reference_wrapper<T>>` to support const/non-const propagation in `cast_op`. (#2705) * Allow type_caster of std::reference_wrapper<T> to be the same as a native reference. Before, both std::reference_wrapper<T> and std::reference_wrapper<const T> would invoke cast_op<type>. This doesn't allow the type_caster<> specialization for T to distinguish reference_wrapper types from value types. After, the type_caster<> specialization invokes cast_op<type&>, which allows reference_wrapper to behave in the same way as a native reference type. * Add tests/examples for std::reference_wrapper<const T> * Add tests which use mutable/immutable variants This test is a chimera; it blends the pybind11 casters with a custom pytype implementation that supports immutable and mutable calls. In order to detect the immutable/mutable state, the cast_op needs to propagate it, even through e.g. std::reference<const T> Note: This is still a work in progress; some things are crashing, which likely means that I have a refcounting bug or something else missing. * Add/finish tests that distinguish const& from & Fixes the bugs in my custom python type implementation, demonstrate test that requires const& and reference_wrapper<const T> being treated differently from Non-const. * Add passing a const to non-const method. * Demonstrate non-const conversion of reference_wrapper in tests. Apply formatting presubmit check. * Fix build errors from presubmit checks. * Try and fix a few more CI errors * More CI fixes. * More CI fixups. * Try and get PyPy to work. * Additional minor fixups. Getting close to CI green. * More ci fixes? * fix clang-tidy warnings from presubmit * fix more clang-tidy warnings * minor comment and consistency cleanups * PyDECREF -> Py_DECREF * copy/move constructors * Resolve codereview comments * more review comment fixes * review comments: remove spurious & * Make the test fail even when the static_assert is commented out. This expands the test_freezable_type_caster a bit by: 1/ adding accessors .is_immutable and .addr to compare identity from python. 2/ Changing the default cast_op of the type_caster<> specialization to return a non-const value. In normal codepaths this is a reasonable default. 3/ adding roundtrip variants to exercise the by reference, by pointer and by reference_wrapper in all call paths. In conjunction with 2/, this demonstrates the failure case of the existing std::reference_wrpper conversion, which now loses const in a similar way that happens when using the default cast_op_type<>. * apply presubmit formatting * Revert inclusion of test_freezable_type_caster There's some concern that this test is a bit unwieldly because of the use of the raw <Python.h> functions. Removing for now. * Add a test that validates const references propagation. This test verifies that cast_op may be used to correctly detect const reference types when used with std::reference_wrapper. * mend * Review comments based changes. 1. std::add_lvalue_reference<type> -> type& 2. Simplify the test a little more; we're never returning the ConstRefCaster type so the class_ definition can be removed. * formatted files again. * Move const_ref_caster test to builtin_casters * Review comments: use cast_op and adjust some comments. * Simplify ConstRefCasted test I like this version better as it moves the assertion that matters back into python. ci: drop pypy2 linux, PGI 20.7, add Python 10 dev (#2724) * ci: drop pypy2 linux, add Python 10 dev * ci: fix mistake * ci: commented-out PGI 20.11, drop 20.7 fix: regression with installed pybind11 overriding local one (#2716) * fix: regression with installed pybind11 overriding discovered one Closes #2709 * docs: wording incorrect style: remove redundant instance->owned = true (#2723) which was just before set to True in instance->allocate_layout() fix: also throw in the move-constructor added by the PYBIND11_OBJECT macro, after the argument has been moved-out (if necessary) (#2701) Make args_are_all_* ICC workarounds unconditional Disable test_aligned on Intel ICC Fix test_aligned on Intel ICC Skip test_python_alreadyset_in_destructor on Intel ICC Fix test_aligned again ICC CI: Downgrade pytest pytest 6 does not capture the `discard_as_unraisable` stderr and just writes a warning with its content instead. * refactor: simpler Intel workaround, suggested by @laramiel * fix: try version with impl to see if it is easier to compile * docs: update README for ICC Co-authored-by: Axel Huebl <axel.huebl@plasma.ninja> Co-authored-by: Henry Schreiner <henryschreineriii@gmail.com>
2021-01-18 00:53:07 +00:00
#endif
#include <pybind11/stl.h>
#include "constructor_stats.h"
#include "local_bindings.h"
#include "pybind11_tests.h"
#include <utility>
#if defined(_MSC_VER)
# pragma warning(disable : 4324)
// warning C4324: structure was padded due to alignment specifier
#endif
// test_brace_initialization
struct NoBraceInitialization {
explicit NoBraceInitialization(std::vector<int> v) : vec{std::move(v)} {}
template <typename T>
NoBraceInitialization(std::initializer_list<T> l) : vec(l) {}
std::vector<int> vec;
};
Reproducer and fix for issue encountered in smart_holder update. (#4228) * Reproducer for issue encountered in smart_holder update. * clang-tidy compatibility (untested). * Add `enable_if_t` to workaround. * Bug fix: Move `PYBIND11_USING_WORKAROUND_FOR_CUDA_11_4_THROUGH_8` determination to detail/common.h So that it actually is defined in pybind11.h * Try using the workaround (which is nicer than the original code) universally. * Reduce reproducer for CUDA 11.7 issue encountered in smart_holder update. This commit tested in isolation on top of current master + first version of reproducer (62311eb431849d135a5db84f6c75ec390f2ede7c). Succeeds with Debian Clang 14.0.6 C++17 (and probably all other compilers). Fails for CUDA 11.7: ``` cd /build/tests && /usr/local/cuda/bin/nvcc -forward-unknown-to-host-compiler -Dpybind11_tests_EXPORTS -I/mounted_pybind11/include -isystem=/usr/include/python3.10 -g --generate-code=arch=compute_52,code=[compute_52,sm_52] -Xcompiler=-fPIC -Xcompiler=-fvisibility=hidden -Werror all-warnings -std=c++17 -MD -MT tests/CMakeFiles/pybind11_tests.dir/test_class.cpp.o -MF CMakeFiles/pybind11_tests.dir/test_class.cpp.o.d -x cu -c /mounted_pybind11/tests/test_class.cpp -o CMakeFiles/pybind11_tests.dir/test_class.cpp.o /mounted_pybind11/tests/test_class.cpp(53): error: more than one instance of overloaded function "pybind11::class_<type_, options...>::def [with type_=test_class::pr4220_tripped_over_this::Empty0, options=<>]" matches the argument list: function template "pybind11::class_<test_class::pr4220_tripped_over_this::Empty0> &pybind11::class_<type_, options...>::def(const char *, Func &&, const Extra &...) [with type_=test_class::pr4220_tripped_over_this::Empty0, options=<>]" /mounted_pybind11/include/pybind11/pybind11.h(1557): here function template "pybind11::class_<test_class::pr4220_tripped_over_this::Empty0> &pybind11::class_<type_, options...>::def(const T &, const Extra &...) [with type_=test_class::pr4220_tripped_over_this::Empty0, options=<>]" /mounted_pybind11/include/pybind11/pybind11.h(1586): here argument types are: (const char [8], <unknown-type>) object type is: pybind11::class_<test_class::pr4220_tripped_over_this::Empty0> 1 error detected in the compilation of "/mounted_pybind11/tests/test_class.cpp". ```
2022-10-10 04:50:35 +00:00
namespace test_class {
namespace pr4220_tripped_over_this { // PR #4227
template <int>
struct SoEmpty {};
template <typename T>
std::string get_msg(const T &) {
return "This is really only meant to exercise successful compilation.";
}
using Empty0 = SoEmpty<0x0>;
void bind_empty0(py::module_ &m) {
py::class_<Empty0>(m, "Empty0").def(py::init<>()).def("get_msg", get_msg<Empty0>);
}
} // namespace pr4220_tripped_over_this
} // namespace test_class
TEST_SUBMODULE(class_, m) {
// test_instance
struct NoConstructor {
NoConstructor() = default;
NoConstructor(const NoConstructor &) = default;
NoConstructor(NoConstructor &&) = default;
static NoConstructor *new_instance() {
auto *ptr = new NoConstructor();
print_created(ptr, "via new_instance");
return ptr;
}
~NoConstructor() { print_destroyed(this); }
};
struct NoConstructorNew {
NoConstructorNew() = default;
NoConstructorNew(const NoConstructorNew &) = default;
NoConstructorNew(NoConstructorNew &&) = default;
static NoConstructorNew *new_instance() {
auto *ptr = new NoConstructorNew();
print_created(ptr, "via new_instance");
return ptr;
}
~NoConstructorNew() { print_destroyed(this); }
};
py::class_<NoConstructor>(m, "NoConstructor")
.def_static("new_instance", &NoConstructor::new_instance, "Return an instance");
py::class_<NoConstructorNew>(m, "NoConstructorNew")
2021-09-20 20:03:21 +00:00
.def(py::init([](const NoConstructorNew &self) { return self; })) // Need a NOOP __init__
.def_static("__new__",
[](const py::object &) { return NoConstructorNew::new_instance(); });
// test_inheritance
class Pet {
public:
Pet(const std::string &name, const std::string &species)
: m_name(name), m_species(species) {}
std::string name() const { return m_name; }
std::string species() const { return m_species; }
private:
std::string m_name;
std::string m_species;
};
class Dog : public Pet {
public:
explicit Dog(const std::string &name) : Pet(name, "dog") {}
std::string bark() const { return "Woof!"; }
};
class Rabbit : public Pet {
public:
explicit Rabbit(const std::string &name) : Pet(name, "parrot") {}
};
class Hamster : public Pet {
public:
explicit Hamster(const std::string &name) : Pet(name, "rodent") {}
};
class Chimera : public Pet {
Chimera() : Pet("Kimmy", "chimera") {}
};
py::class_<Pet> pet_class(m, "Pet");
pet_class.def(py::init<std::string, std::string>())
.def("name", &Pet::name)
.def("species", &Pet::species);
/* One way of declaring a subclass relationship: reference parent's class_ object */
py::class_<Dog>(m, "Dog", pet_class).def(py::init<std::string>());
/* Another way of declaring a subclass relationship: reference parent's C++ type */
py::class_<Rabbit, Pet>(m, "Rabbit").def(py::init<std::string>());
/* And another: list parent in class template arguments */
py::class_<Hamster, Pet>(m, "Hamster").def(py::init<std::string>());
/* Constructors are not inherited by default */
py::class_<Chimera, Pet>(m, "Chimera");
m.def("pet_name_species",
[](const Pet &pet) { return pet.name() + " is a " + pet.species(); });
m.def("dog_bark", [](const Dog &dog) { return dog.bark(); });
// test_automatic_upcasting
struct BaseClass {
BaseClass() = default;
BaseClass(const BaseClass &) = default;
BaseClass(BaseClass &&) = default;
virtual ~BaseClass() = default;
};
struct DerivedClass1 : BaseClass {};
struct DerivedClass2 : BaseClass {};
py::class_<BaseClass>(m, "BaseClass").def(py::init<>());
py::class_<DerivedClass1>(m, "DerivedClass1").def(py::init<>());
py::class_<DerivedClass2>(m, "DerivedClass2").def(py::init<>());
m.def("return_class_1", []() -> BaseClass * { return new DerivedClass1(); });
m.def("return_class_2", []() -> BaseClass * { return new DerivedClass2(); });
m.def("return_class_n", [](int n) -> BaseClass * {
if (n == 1) {
return new DerivedClass1();
}
if (n == 2) {
return new DerivedClass2();
}
return new BaseClass();
});
m.def("return_none", []() -> BaseClass * { return nullptr; });
// test_isinstance
m.def("check_instances", [](const py::list &l) {
return py::make_tuple(py::isinstance<py::tuple>(l[0]),
py::isinstance<py::dict>(l[1]),
py::isinstance<Pet>(l[2]),
py::isinstance<Pet>(l[3]),
py::isinstance<Dog>(l[4]),
py::isinstance<Rabbit>(l[5]),
py::isinstance<UnregisteredType>(l[6]));
});
struct Invalid {};
// test_type
m.def("check_type", [](int category) {
// Currently not supported (via a fail at compile time)
// See https://github.com/pybind/pybind11/issues/2486
// if (category == 2)
// return py::type::of<int>();
if (category == 1) {
return py::type::of<DerivedClass1>();
}
return py::type::of<Invalid>();
});
m.def("get_type_of", [](py::object ob) { return py::type::of(std::move(ob)); });
m.def("get_type_classic", [](py::handle h) { return h.get_type(); });
m.def("as_type", [](const py::object &ob) { return py::type(ob); });
// test_mismatched_holder
struct MismatchBase1 {};
struct MismatchDerived1 : MismatchBase1 {};
struct MismatchBase2 {};
struct MismatchDerived2 : MismatchBase2 {};
m.def("mismatched_holder_1", []() {
auto mod = py::module_::import("__main__");
py::class_<MismatchBase1, std::shared_ptr<MismatchBase1>>(mod, "MismatchBase1");
py::class_<MismatchDerived1, MismatchBase1>(mod, "MismatchDerived1");
});
m.def("mismatched_holder_2", []() {
auto mod = py::module_::import("__main__");
py::class_<MismatchBase2>(mod, "MismatchBase2");
py::class_<MismatchDerived2, std::shared_ptr<MismatchDerived2>, MismatchBase2>(
mod, "MismatchDerived2");
});
// test_override_static
// #511: problem with inheritance + overwritten def_static
struct MyBase {
static std::unique_ptr<MyBase> make() { return std::unique_ptr<MyBase>(new MyBase()); }
};
struct MyDerived : MyBase {
static std::unique_ptr<MyDerived> make() {
return std::unique_ptr<MyDerived>(new MyDerived());
}
};
py::class_<MyBase>(m, "MyBase").def_static("make", &MyBase::make);
py::class_<MyDerived, MyBase>(m, "MyDerived")
.def_static("make", &MyDerived::make)
.def_static("make2", &MyDerived::make);
// test_implicit_conversion_life_support
struct ConvertibleFromUserType {
int i;
explicit ConvertibleFromUserType(UserType u) : i(u.value()) {}
};
py::class_<ConvertibleFromUserType>(m, "AcceptsUserType").def(py::init<UserType>());
py::implicitly_convertible<UserType, ConvertibleFromUserType>();
m.def("implicitly_convert_argument", [](const ConvertibleFromUserType &r) { return r.i; });
m.def("implicitly_convert_variable", [](const py::object &o) {
// `o` is `UserType` and `r` is a reference to a temporary created by implicit
// conversion. This is valid when called inside a bound function because the temp
// object is attached to the same life support system as the arguments.
const auto &r = o.cast<const ConvertibleFromUserType &>();
return r.i;
});
m.add_object("implicitly_convert_variable_fail", [&] {
auto f = [](PyObject *, PyObject *args) -> PyObject * {
auto o = py::reinterpret_borrow<py::tuple>(args)[0];
try { // It should fail here because there is no life support.
o.cast<const ConvertibleFromUserType &>();
} catch (const py::cast_error &e) {
return py::str(e.what()).release().ptr();
}
return py::str().release().ptr();
};
auto *def = new PyMethodDef{"f", f, METH_VARARGS, nullptr};
py::capsule def_capsule(def,
[](void *ptr) { delete reinterpret_cast<PyMethodDef *>(ptr); });
return py::reinterpret_steal<py::object>(
PyCFunction_NewEx(def, def_capsule.ptr(), m.ptr()));
}());
// test_operator_new_delete
struct HasOpNewDel {
std::uint64_t i;
static void *operator new(size_t s) {
py::print("A new", s);
return ::operator new(s);
}
static void *operator new(size_t s, void *ptr) {
py::print("A placement-new", s);
return ptr;
}
static void operator delete(void *p) {
py::print("A delete");
return ::operator delete(p);
}
};
struct HasOpNewDelSize {
std::uint32_t i;
static void *operator new(size_t s) {
py::print("B new", s);
return ::operator new(s);
}
static void *operator new(size_t s, void *ptr) {
py::print("B placement-new", s);
return ptr;
}
static void operator delete(void *p, size_t s) {
py::print("B delete", s);
return ::operator delete(p);
}
};
struct AliasedHasOpNewDelSize {
std::uint64_t i;
static void *operator new(size_t s) {
py::print("C new", s);
return ::operator new(s);
}
static void *operator new(size_t s, void *ptr) {
py::print("C placement-new", s);
return ptr;
}
static void operator delete(void *p, size_t s) {
py::print("C delete", s);
return ::operator delete(p);
}
virtual ~AliasedHasOpNewDelSize() = default;
2020-07-24 01:16:54 +00:00
AliasedHasOpNewDelSize() = default;
AliasedHasOpNewDelSize(const AliasedHasOpNewDelSize &) = delete;
};
struct PyAliasedHasOpNewDelSize : AliasedHasOpNewDelSize {
PyAliasedHasOpNewDelSize() = default;
explicit PyAliasedHasOpNewDelSize(int) {}
std::uint64_t j;
};
struct HasOpNewDelBoth {
std::uint32_t i[8];
static void *operator new(size_t s) {
py::print("D new", s);
return ::operator new(s);
}
static void *operator new(size_t s, void *ptr) {
py::print("D placement-new", s);
return ptr;
}
static void operator delete(void *p) {
py::print("D delete");
return ::operator delete(p);
}
static void operator delete(void *p, size_t s) {
py::print("D wrong delete", s);
return ::operator delete(p);
}
};
py::class_<HasOpNewDel>(m, "HasOpNewDel").def(py::init<>());
py::class_<HasOpNewDelSize>(m, "HasOpNewDelSize").def(py::init<>());
py::class_<HasOpNewDelBoth>(m, "HasOpNewDelBoth").def(py::init<>());
py::class_<AliasedHasOpNewDelSize, PyAliasedHasOpNewDelSize> aliased(m,
"AliasedHasOpNewDelSize");
aliased.def(py::init<>());
aliased.attr("size_noalias") = py::int_(sizeof(AliasedHasOpNewDelSize));
aliased.attr("size_alias") = py::int_(sizeof(PyAliasedHasOpNewDelSize));
// This test is actually part of test_local_bindings (test_duplicate_local), but we need a
// definition in a different compilation unit within the same module:
bind_local<LocalExternal, 17>(m, "LocalExternal", py::module_local());
// test_bind_protected_functions
class ProtectedA {
protected:
int foo() const { return value; }
private:
int value = 42;
};
class PublicistA : public ProtectedA {
public:
using ProtectedA::foo;
};
Dropping MSVC 2015 (#3722) * Changing `_MSC_VER` guard to `< 1910` (dropping MSVC 2015). * Removing MSVC 2015 from ci.yml, and .appveyor.yml entirely. * Bringing back .appveyor.yml from master. * appveyor Visual Studio 2017 * 1st manual pass, builds & tests with unix_clang, before pre-commit. * After clang-format (via pre-commit). * Manual pass looking for "2015", builds & tests with unix_clang, before pre-commit. * Backtracking for include/pybind11 changes in previous commit. git checkout d07865846c7d31dd61111e6df801864327b65070 include/pybind11/attr.h include/pybind11/detail/common.h include/pybind11/functional.h -------------------- CI #4160 errors observed: https://github.com/pybind/pybind11/pull/3722/commits/2a26873727214c5f1e159cba98f5c625b908381a https://github.com/pybind/pybind11/runs/5168332130?check_suite_focus=true $ grep ' error C' *.txt | sed 's/2022-02-12[^ ]*//' | sed 's/^[0-9][0-9]*//' | sed 's/^.*\.txt: //' | sort | uniqD:\a\pybind11\pybind11\include\pybind11\cast.h(1364,1): error C2752: 'pybind11::detail::type_caster<Eigen::Ref<Eigen::Vector3f,0,pybind11::EigenDStride>,void>': more than one partial specialization matches the template argument list [D:\a\pybind11\pybind11\build\tests\pybind11_tests.vcxproj] d:\a\pybind11\pybind11\include\pybind11\detail/common.h(1023): error C2737: 'pybind11::overload_cast': 'constexpr' object must be initialized [D:\a\pybind11\pybind11\build\tests\cross_module_gil_utils.vcxproj] d:\a\pybind11\pybind11\include\pybind11\detail/common.h(1023): error C2737: 'pybind11::overload_cast': 'constexpr' object must be initialized [D:\a\pybind11\pybind11\build\tests\pybind11_cross_module_tests.vcxproj] d:\a\pybind11\pybind11\include\pybind11\detail/common.h(1023): error C2737: 'pybind11::overload_cast': 'constexpr' object must be initialized [D:\a\pybind11\pybind11\build\tests\pybind11_tests.vcxproj] d:\a\pybind11\pybind11\include\pybind11\detail/common.h(1023): error C2737: 'pybind11::overload_cast': 'constexpr' object must be initialized [D:\a\pybind11\pybind11\build\tests\test_embed\external_module.vcxproj] D:\a\pybind11\pybind11\include\pybind11\detail/common.h(624): fatal error C1001: Internal compiler error. [D:\a\pybind11\pybind11\build\tests\pybind11_tests.vcxproj] D:\a\pybind11\pybind11\include\pybind11\detail/common.h(624): fatal error C1001: Internal compiler error. [D:\a\pybind11\pybind11\tests\pybind11_tests.vcxproj] $ grep ': error C2737' *.txt | sed 's/^.*MSVC//' | sed 's/___.*//' | sort | uniq _2017 $ grep ': error C2752' *.txt 3______3.8_____MSVC_2019_____x86_-DCMAKE_CXX_STANDARD=17.txt:2022-02-12T16:12:45.9921122Z D:\a\pybind11\pybind11\include\pybind11\cast.h(1364,1): error C2752: 'pybind11::detail::type_caster<Eigen::Ref<Eigen::Vector3f,0,pybind11::EigenDStride>,void>': more than one partial specialization matches the template argument list [D:\a\pybind11\pybind11\build\tests\pybind11_tests.vcxproj] $ grep ': fatal error C1001:' *.txt 10______pypy-3.8-v7.3.7_____windows-2022_____x64.txt:2022-02-12T16:12:56.3163683Z D:\a\pybind11\pybind11\include\pybind11\detail/common.h(624): fatal error C1001: Internal compiler error. [D:\a\pybind11\pybind11\tests\pybind11_tests.vcxproj] 1______3.6_____MSVC_2019_____x86.txt:2022-02-12T16:12:47.6774625Z D:\a\pybind11\pybind11\include\pybind11\detail/common.h(624): fatal error C1001: Internal compiler error. [D:\a\pybind11\pybind11\build\tests\pybind11_tests.vcxproj] 16______3.6_____windows-latest_____x64_-DPYBIND11_FINDPYTHON=ON.txt:2022-02-12T16:12:27.0556151Z D:\a\pybind11\pybind11\include\pybind11\detail/common.h(624): fatal error C1001: Internal compiler error. [D:\a\pybind11\pybind11\tests\pybind11_tests.vcxproj] 17______3.9_____windows-2019_____x64.txt:2022-02-12T16:12:30.3822566Z D:\a\pybind11\pybind11\include\pybind11\detail/common.h(624): fatal error C1001: Internal compiler error. [D:\a\pybind11\pybind11\tests\pybind11_tests.vcxproj] 2______3.7_____MSVC_2019_____x86.txt:2022-02-12T16:12:38.7018911Z D:\a\pybind11\pybind11\include\pybind11\detail/common.h(624): fatal error C1001: Internal compiler error. [D:\a\pybind11\pybind11\build\tests\pybind11_tests.vcxproj] 6______3.6_____windows-2022_____x64.txt:2022-02-12T16:12:00.4513642Z D:\a\pybind11\pybind11\include\pybind11\detail/common.h(624): fatal error C1001: Internal compiler error. [D:\a\pybind11\pybind11\tests\pybind11_tests.vcxproj] 7______3.9_____windows-2022_____x64.txt:2022-02-12T16:11:43.6306160Z D:\a\pybind11\pybind11\include\pybind11\detail/common.h(624): fatal error C1001: Internal compiler error. [D:\a\pybind11\pybind11\tests\pybind11_tests.vcxproj] 8______3.10_____windows-2022_____x64.txt:2022-02-12T16:11:49.9589644Z D:\a\pybind11\pybind11\include\pybind11\detail/common.h(624): fatal error C1001: Internal compiler error. [D:\a\pybind11\pybind11\tests\pybind11_tests.vcxproj] 9______pypy-3.7-v7.3.7_____windows-2022_____x64.txt:2022-02-12T16:11:53.7912112Z D:\a\pybind11\pybind11\include\pybind11\detail/common.h(624): fatal error C1001: Internal compiler error. [D:\a\pybind11\pybind11\tests\pybind11_tests.vcxproj] * common.h: is_template_base_of * Re-applying 4 changes from 2a26873727214c5f1e159cba98f5c625b908381a that work universally. * `overload_cast = {};` only for MSVC 2017 and Clang 5 * Refining condition for using is_template_base_of workaround. * Undoing MSVC 2015 workaround in test_constants_and_functions.cpp * CentOS7: silence_unused_warnings * Tweaks in response to reviews. * Adding windows-2022 C++20 * Trying another way of adding windows-2022 C++20
2022-02-14 19:36:22 +00:00
py::class_<ProtectedA>(m, "ProtectedA").def(py::init<>()).def("foo", &PublicistA::foo);
class ProtectedB {
public:
virtual ~ProtectedB() = default;
2020-07-24 01:16:54 +00:00
ProtectedB() = default;
ProtectedB(const ProtectedB &) = delete;
protected:
virtual int foo() const { return value; }
private:
int value = 42;
};
class TrampolineB : public ProtectedB {
public:
int foo() const override { PYBIND11_OVERRIDE(int, ProtectedB, foo, ); }
};
class PublicistB : public ProtectedB {
public:
ci: Intel icc/icpc via oneAPI (#2573) * CI: Intel icc/icpc via oneAPI Add testing for Intel icc/icpc via the oneAPI images. Intel oneAPI is in a late beta stage, currently shipping oneAPI beta09 with ICC 20.2. * CI: Skip Interpreter Tests for Intel Cannot find how to add this, neiter the package `libc6-dev` nor `intel-oneapi-mkl-devel` help when installed to solve this: ``` -- Looking for C++ include pthread.h -- Looking for C++ include pthread.h - not found CMake Error at /__t/cmake/3.18.4/x64/cmake-3.18.4-Linux-x86_64/share/cmake-3.18/Modules/FindPackageHandleStandardArgs.cmake:165 (message): Could NOT find Threads (missing: Threads_FOUND) Call Stack (most recent call first): /__t/cmake/3.18.4/x64/cmake-3.18.4-Linux-x86_64/share/cmake-3.18/Modules/FindPackageHandleStandardArgs.cmake:458 (_FPHSA_FAILURE_MESSAGE) /__t/cmake/3.18.4/x64/cmake-3.18.4-Linux-x86_64/share/cmake-3.18/Modules/FindThreads.cmake:234 (FIND_PACKAGE_HANDLE_STANDARD_ARGS) tests/test_embed/CMakeLists.txt:17 (find_package) ``` * CI: libc6-dev from GCC for ICC * CI: Run bare metal for oneAPI * CI: Ubuntu 18.04 for oneAPI * CI: Intel +Catch -Eigen * CI: CMake from Apt (ICC tests) * CI: Replace Intel Py with GCC Py * CI: Intel w/o GCC's Eigen * CI: ICC with verbose make * [Debug] Find core dump * tests: use arg{} instead of arg() for Intel * tests: adding a few more missing {} * fix: sync with @tobiasleibner's branch * fix: try ubuntu 20-04 * fix: drop exit 1 * style: clang tidy fix * style: fix missing NOLINT * ICC: Update Compiler Name Changed upstream with the last oneAPI release. * ICC CI: Downgrade pytest pytest 6 does not capture the `discard_as_unraisable` stderr and just writes a warning with its content instead. * Use new test pinning requirements.txt * tests: add notes about intel, cleanup Co-authored-by: Henry Schreiner <henryschreineriii@gmail.com>
2021-01-15 20:59:47 +00:00
// [workaround(intel)] = default does not work here
fix: Intel ICC C++17 compatibility (#2729) * CI: Intel icc/icpc via oneAPI Add testing for Intel icc/icpc via the oneAPI images. Intel oneAPI is in a late beta stage, currently shipping oneAPI beta09 with ICC 20.2. CI: Skip Interpreter Tests for Intel Cannot find how to add this, neiter the package `libc6-dev` nor `intel-oneapi-mkl-devel` help when installed to solve this: ``` -- Looking for C++ include pthread.h -- Looking for C++ include pthread.h - not found CMake Error at /__t/cmake/3.18.4/x64/cmake-3.18.4-Linux-x86_64/share/cmake-3.18/Modules/FindPackageHandleStandardArgs.cmake:165 (message): Could NOT find Threads (missing: Threads_FOUND) Call Stack (most recent call first): /__t/cmake/3.18.4/x64/cmake-3.18.4-Linux-x86_64/share/cmake-3.18/Modules/FindPackageHandleStandardArgs.cmake:458 (_FPHSA_FAILURE_MESSAGE) /__t/cmake/3.18.4/x64/cmake-3.18.4-Linux-x86_64/share/cmake-3.18/Modules/FindThreads.cmake:234 (FIND_PACKAGE_HANDLE_STANDARD_ARGS) tests/test_embed/CMakeLists.txt:17 (find_package) ``` CI: libc6-dev from GCC for ICC CI: Run bare metal for oneAPI CI: Ubuntu 18.04 for oneAPI CI: Intel +Catch -Eigen CI: CMake from Apt (ICC tests) CI: Replace Intel Py with GCC Py CI: Intel w/o GCC's Eigen CI: ICC with verbose make [Debug] Find core dump tests: use arg{} instead of arg() for Intel tests: adding a few more missing {} fix: sync with @tobiasleibner's branch fix: try ubuntu 20-04 fix: drop exit 1 docs: Apply suggestions from code review Co-authored-by: Tobias Leibner <tobias.leibner@googlemail.com> Workaround for ICC enable_if issues Another workaround for ICC's enable_if issues fix error in previous commit Disable one test for the Intel compiler in C++17 mode Add back one instance of py::arg().noconvert() Add NOLINT to fix clang-tidy check Work around for ICC internal error in PYBIND11_EXPAND_SIDE_EFFECTS in C++17 mode CI: Intel ICC with C++17 docs: pybind11/numpy.h does not require numpy at build time. (#2720) This is nice enough to be mentioned explicitly in the docs. docs: Update warning about Python 3.9.0 UB, now that 3.9.1 has been released (#2719) Adjusting `type_caster<std::reference_wrapper<T>>` to support const/non-const propagation in `cast_op`. (#2705) * Allow type_caster of std::reference_wrapper<T> to be the same as a native reference. Before, both std::reference_wrapper<T> and std::reference_wrapper<const T> would invoke cast_op<type>. This doesn't allow the type_caster<> specialization for T to distinguish reference_wrapper types from value types. After, the type_caster<> specialization invokes cast_op<type&>, which allows reference_wrapper to behave in the same way as a native reference type. * Add tests/examples for std::reference_wrapper<const T> * Add tests which use mutable/immutable variants This test is a chimera; it blends the pybind11 casters with a custom pytype implementation that supports immutable and mutable calls. In order to detect the immutable/mutable state, the cast_op needs to propagate it, even through e.g. std::reference<const T> Note: This is still a work in progress; some things are crashing, which likely means that I have a refcounting bug or something else missing. * Add/finish tests that distinguish const& from & Fixes the bugs in my custom python type implementation, demonstrate test that requires const& and reference_wrapper<const T> being treated differently from Non-const. * Add passing a const to non-const method. * Demonstrate non-const conversion of reference_wrapper in tests. Apply formatting presubmit check. * Fix build errors from presubmit checks. * Try and fix a few more CI errors * More CI fixes. * More CI fixups. * Try and get PyPy to work. * Additional minor fixups. Getting close to CI green. * More ci fixes? * fix clang-tidy warnings from presubmit * fix more clang-tidy warnings * minor comment and consistency cleanups * PyDECREF -> Py_DECREF * copy/move constructors * Resolve codereview comments * more review comment fixes * review comments: remove spurious & * Make the test fail even when the static_assert is commented out. This expands the test_freezable_type_caster a bit by: 1/ adding accessors .is_immutable and .addr to compare identity from python. 2/ Changing the default cast_op of the type_caster<> specialization to return a non-const value. In normal codepaths this is a reasonable default. 3/ adding roundtrip variants to exercise the by reference, by pointer and by reference_wrapper in all call paths. In conjunction with 2/, this demonstrates the failure case of the existing std::reference_wrpper conversion, which now loses const in a similar way that happens when using the default cast_op_type<>. * apply presubmit formatting * Revert inclusion of test_freezable_type_caster There's some concern that this test is a bit unwieldly because of the use of the raw <Python.h> functions. Removing for now. * Add a test that validates const references propagation. This test verifies that cast_op may be used to correctly detect const reference types when used with std::reference_wrapper. * mend * Review comments based changes. 1. std::add_lvalue_reference<type> -> type& 2. Simplify the test a little more; we're never returning the ConstRefCaster type so the class_ definition can be removed. * formatted files again. * Move const_ref_caster test to builtin_casters * Review comments: use cast_op and adjust some comments. * Simplify ConstRefCasted test I like this version better as it moves the assertion that matters back into python. ci: drop pypy2 linux, PGI 20.7, add Python 10 dev (#2724) * ci: drop pypy2 linux, add Python 10 dev * ci: fix mistake * ci: commented-out PGI 20.11, drop 20.7 fix: regression with installed pybind11 overriding local one (#2716) * fix: regression with installed pybind11 overriding discovered one Closes #2709 * docs: wording incorrect style: remove redundant instance->owned = true (#2723) which was just before set to True in instance->allocate_layout() fix: also throw in the move-constructor added by the PYBIND11_OBJECT macro, after the argument has been moved-out (if necessary) (#2701) Make args_are_all_* ICC workarounds unconditional Disable test_aligned on Intel ICC Fix test_aligned on Intel ICC Skip test_python_alreadyset_in_destructor on Intel ICC Fix test_aligned again ICC CI: Downgrade pytest pytest 6 does not capture the `discard_as_unraisable` stderr and just writes a warning with its content instead. * refactor: simpler Intel workaround, suggested by @laramiel * fix: try version with impl to see if it is easier to compile * docs: update README for ICC Co-authored-by: Axel Huebl <axel.huebl@plasma.ninja> Co-authored-by: Henry Schreiner <henryschreineriii@gmail.com>
2021-01-18 00:53:07 +00:00
// Removing or defaulting this destructor results in linking errors with the Intel compiler
// (in Debug builds only, tested with icpc (ICC) 2021.1 Beta 20200827)
~PublicistB() override{}; // NOLINT(modernize-use-equals-default)
using ProtectedB::foo;
};
py::class_<ProtectedB, TrampolineB>(m, "ProtectedB")
.def(py::init<>())
.def("foo", &PublicistB::foo);
2017-08-21 20:48:28 +00:00
// test_brace_initialization
struct BraceInitialization {
int field1;
std::string field2;
};
py::class_<BraceInitialization>(m, "BraceInitialization")
.def(py::init<int, const std::string &>())
.def_readwrite("field1", &BraceInitialization::field1)
.def_readwrite("field2", &BraceInitialization::field2);
// We *don't* want to construct using braces when the given constructor argument maps to a
// constructor, because brace initialization could go to the wrong place (in particular when
// there is also an `initializer_list<T>`-accept constructor):
py::class_<NoBraceInitialization>(m, "NoBraceInitialization")
.def(py::init<std::vector<int>>())
.def_readonly("vec", &NoBraceInitialization::vec);
// test_reentrant_implicit_conversion_failure
// #1035: issue with runaway reentrant implicit conversion
struct BogusImplicitConversion {
BogusImplicitConversion(const BogusImplicitConversion &) = default;
};
py::class_<BogusImplicitConversion>(m, "BogusImplicitConversion")
.def(py::init<const BogusImplicitConversion &>());
py::implicitly_convertible<int, BogusImplicitConversion>();
// test_qualname
// #1166: nested class docstring doesn't show nested name
// Also related: tests that __qualname__ is set properly
struct NestBase {};
struct Nested {};
py::class_<NestBase> base(m, "NestBase");
base.def(py::init<>());
py::class_<Nested>(base, "Nested")
.def(py::init<>())
.def("fn", [](Nested &, int, NestBase &, Nested &) {})
.def(
"fa", [](Nested &, int, NestBase &, Nested &) {}, "a"_a, "b"_a, "c"_a);
base.def("g", [](NestBase &, Nested &) {});
base.def("h", []() { return NestBase(); });
// test_error_after_conversion
// The second-pass path through dispatcher() previously didn't
// remember which overload was used, and would crash trying to
// generate a useful error message
struct NotRegistered {};
struct StringWrapper {
std::string str;
};
m.def("test_error_after_conversions", [](int) {});
m.def("test_error_after_conversions",
[](const StringWrapper &) -> NotRegistered { return {}; });
py::class_<StringWrapper>(m, "StringWrapper").def(py::init<std::string>());
py::implicitly_convertible<std::string, StringWrapper>();
#if defined(PYBIND11_CPP17)
struct alignas(1024) Aligned {
std::uintptr_t ptr() const { return (uintptr_t) this; }
};
py::class_<Aligned>(m, "Aligned").def(py::init<>()).def("ptr", &Aligned::ptr);
#endif
// test_final
struct IsFinal final {};
py::class_<IsFinal>(m, "IsFinal", py::is_final());
// test_non_final_final
struct IsNonFinalFinal {};
py::class_<IsNonFinalFinal>(m, "IsNonFinalFinal", py::is_final());
// test_exception_rvalue_abort
struct PyPrintDestructor {
PyPrintDestructor() = default;
~PyPrintDestructor() { py::print("Print from destructor"); }
void throw_something() { throw std::runtime_error("error"); }
};
py::class_<PyPrintDestructor>(m, "PyPrintDestructor")
.def(py::init<>())
.def("throw_something", &PyPrintDestructor::throw_something);
// test_multiple_instances_with_same_pointer
struct SamePointer {};
static SamePointer samePointer;
py::class_<SamePointer, std::unique_ptr<SamePointer, py::nodelete>>(m, "SamePointer")
.def(py::init([]() { return &samePointer; }));
struct Empty {};
py::class_<Empty>(m, "Empty").def(py::init<>());
// test_base_and_derived_nested_scope
struct BaseWithNested {
struct Nested {};
};
struct DerivedWithNested : BaseWithNested {
struct Nested {};
};
py::class_<BaseWithNested> baseWithNested_class(m, "BaseWithNested");
py::class_<DerivedWithNested, BaseWithNested> derivedWithNested_class(m, "DerivedWithNested");
py::class_<BaseWithNested::Nested>(baseWithNested_class, "Nested")
.def_static("get_name", []() { return "BaseWithNested::Nested"; });
py::class_<DerivedWithNested::Nested>(derivedWithNested_class, "Nested")
.def_static("get_name", []() { return "DerivedWithNested::Nested"; });
// test_register_duplicate_class
struct Duplicate {};
struct OtherDuplicate {};
struct DuplicateNested {};
struct OtherDuplicateNested {};
m.def("register_duplicate_class_name", [](const py::module_ &m) {
py::class_<Duplicate>(m, "Duplicate");
py::class_<OtherDuplicate>(m, "Duplicate");
});
m.def("register_duplicate_class_type", [](const py::module_ &m) {
py::class_<OtherDuplicate>(m, "OtherDuplicate");
py::class_<OtherDuplicate>(m, "YetAnotherDuplicate");
});
m.def("register_duplicate_nested_class_name", [](const py::object &gt) {
py::class_<DuplicateNested>(gt, "DuplicateNested");
py::class_<OtherDuplicateNested>(gt, "DuplicateNested");
});
m.def("register_duplicate_nested_class_type", [](const py::object &gt) {
py::class_<OtherDuplicateNested>(gt, "OtherDuplicateNested");
py::class_<OtherDuplicateNested>(gt, "YetAnotherDuplicateNested");
});
Reproducer and fix for issue encountered in smart_holder update. (#4228) * Reproducer for issue encountered in smart_holder update. * clang-tidy compatibility (untested). * Add `enable_if_t` to workaround. * Bug fix: Move `PYBIND11_USING_WORKAROUND_FOR_CUDA_11_4_THROUGH_8` determination to detail/common.h So that it actually is defined in pybind11.h * Try using the workaround (which is nicer than the original code) universally. * Reduce reproducer for CUDA 11.7 issue encountered in smart_holder update. This commit tested in isolation on top of current master + first version of reproducer (62311eb431849d135a5db84f6c75ec390f2ede7c). Succeeds with Debian Clang 14.0.6 C++17 (and probably all other compilers). Fails for CUDA 11.7: ``` cd /build/tests && /usr/local/cuda/bin/nvcc -forward-unknown-to-host-compiler -Dpybind11_tests_EXPORTS -I/mounted_pybind11/include -isystem=/usr/include/python3.10 -g --generate-code=arch=compute_52,code=[compute_52,sm_52] -Xcompiler=-fPIC -Xcompiler=-fvisibility=hidden -Werror all-warnings -std=c++17 -MD -MT tests/CMakeFiles/pybind11_tests.dir/test_class.cpp.o -MF CMakeFiles/pybind11_tests.dir/test_class.cpp.o.d -x cu -c /mounted_pybind11/tests/test_class.cpp -o CMakeFiles/pybind11_tests.dir/test_class.cpp.o /mounted_pybind11/tests/test_class.cpp(53): error: more than one instance of overloaded function "pybind11::class_<type_, options...>::def [with type_=test_class::pr4220_tripped_over_this::Empty0, options=<>]" matches the argument list: function template "pybind11::class_<test_class::pr4220_tripped_over_this::Empty0> &pybind11::class_<type_, options...>::def(const char *, Func &&, const Extra &...) [with type_=test_class::pr4220_tripped_over_this::Empty0, options=<>]" /mounted_pybind11/include/pybind11/pybind11.h(1557): here function template "pybind11::class_<test_class::pr4220_tripped_over_this::Empty0> &pybind11::class_<type_, options...>::def(const T &, const Extra &...) [with type_=test_class::pr4220_tripped_over_this::Empty0, options=<>]" /mounted_pybind11/include/pybind11/pybind11.h(1586): here argument types are: (const char [8], <unknown-type>) object type is: pybind11::class_<test_class::pr4220_tripped_over_this::Empty0> 1 error detected in the compilation of "/mounted_pybind11/tests/test_class.cpp". ```
2022-10-10 04:50:35 +00:00
test_class::pr4220_tripped_over_this::bind_empty0(m);
}
template <int N>
class BreaksBase {
public:
2020-07-24 01:16:54 +00:00
virtual ~BreaksBase() = default;
BreaksBase() = default;
BreaksBase(const BreaksBase &) = delete;
2020-07-24 01:16:54 +00:00
};
template <int N>
class BreaksTramp : public BreaksBase<N> {};
// These should all compile just fine:
2020-09-11 02:57:10 +00:00
using DoesntBreak1 = py::class_<BreaksBase<1>, std::unique_ptr<BreaksBase<1>>, BreaksTramp<1>>;
using DoesntBreak2 = py::class_<BreaksBase<2>, BreaksTramp<2>, std::unique_ptr<BreaksBase<2>>>;
using DoesntBreak3 = py::class_<BreaksBase<3>, std::unique_ptr<BreaksBase<3>>>;
using DoesntBreak4 = py::class_<BreaksBase<4>, BreaksTramp<4>>;
using DoesntBreak5 = py::class_<BreaksBase<5>>;
using DoesntBreak6 = py::class_<BreaksBase<6>, std::shared_ptr<BreaksBase<6>>, BreaksTramp<6>>;
using DoesntBreak7 = py::class_<BreaksBase<7>, BreaksTramp<7>, std::shared_ptr<BreaksBase<7>>>;
using DoesntBreak8 = py::class_<BreaksBase<8>, std::shared_ptr<BreaksBase<8>>>;
#define CHECK_BASE(N) \
static_assert(std::is_same<typename DoesntBreak##N::type, BreaksBase<(N)>>::value, \
"DoesntBreak" #N " has wrong type!")
CHECK_BASE(1);
CHECK_BASE(2);
CHECK_BASE(3);
CHECK_BASE(4);
CHECK_BASE(5);
CHECK_BASE(6);
CHECK_BASE(7);
CHECK_BASE(8);
#define CHECK_ALIAS(N) \
static_assert( \
DoesntBreak##N::has_alias \
&& std::is_same<typename DoesntBreak##N::type_alias, BreaksTramp<(N)>>::value, \
"DoesntBreak" #N " has wrong type_alias!")
#define CHECK_NOALIAS(N) \
static_assert(!DoesntBreak##N::has_alias \
&& std::is_void<typename DoesntBreak##N::type_alias>::value, \
"DoesntBreak" #N " has type alias, but shouldn't!")
CHECK_ALIAS(1);
CHECK_ALIAS(2);
CHECK_NOALIAS(3);
CHECK_ALIAS(4);
CHECK_NOALIAS(5);
CHECK_ALIAS(6);
CHECK_ALIAS(7);
CHECK_NOALIAS(8);
#define CHECK_HOLDER(N, TYPE) \
static_assert(std::is_same<typename DoesntBreak##N::holder_type, \
std::TYPE##_ptr<BreaksBase<(N)>>>::value, \
"DoesntBreak" #N " has wrong holder_type!")
CHECK_HOLDER(1, unique);
CHECK_HOLDER(2, unique);
CHECK_HOLDER(3, unique);
CHECK_HOLDER(4, unique);
CHECK_HOLDER(5, unique);
CHECK_HOLDER(6, shared);
CHECK_HOLDER(7, shared);
CHECK_HOLDER(8, shared);
// There's no nice way to test that these fail because they fail to compile; leave them here,
// though, so that they can be manually tested by uncommenting them (and seeing that compilation
// failures occurs).
// We have to actually look into the type: the typedef alone isn't enough to instantiate the type:
#define CHECK_BROKEN(N) \
static_assert(std::is_same<typename Breaks##N::type, BreaksBase<-(N)>>::value, \
"Breaks1 has wrong type!");
#ifdef PYBIND11_NEVER_DEFINED_EVER
// Two holder classes:
typedef py::
class_<BreaksBase<-1>, std::unique_ptr<BreaksBase<-1>>, std::unique_ptr<BreaksBase<-1>>>
Breaks1;
CHECK_BROKEN(1);
// Two aliases:
typedef py::class_<BreaksBase<-2>, BreaksTramp<-2>, BreaksTramp<-2>> Breaks2;
CHECK_BROKEN(2);
// Holder + 2 aliases
typedef py::
class_<BreaksBase<-3>, std::unique_ptr<BreaksBase<-3>>, BreaksTramp<-3>, BreaksTramp<-3>>
Breaks3;
CHECK_BROKEN(3);
// Alias + 2 holders
typedef py::class_<BreaksBase<-4>,
std::unique_ptr<BreaksBase<-4>>,
BreaksTramp<-4>,
std::shared_ptr<BreaksBase<-4>>>
Breaks4;
CHECK_BROKEN(4);
// Invalid option (not a subclass or holder)
typedef py::class_<BreaksBase<-5>, BreaksTramp<-4>> Breaks5;
CHECK_BROKEN(5);
// Invalid option: multiple inheritance not supported:
template <>
struct BreaksBase<-8> : BreaksBase<-6>, BreaksBase<-7> {};
typedef py::class_<BreaksBase<-8>, BreaksBase<-6>, BreaksBase<-7>> Breaks8;
CHECK_BROKEN(8);
#endif