* Unify cast_error message thrown by [simple|unpacking]_collector
simple_collector and unpacking_collector throw different error messages
when the casting of an argument failed: While the former mentions make_tuple(),
the latter emphasises the call argument (and its name/position).
* Consolidating "Unable to convert call argument" error reporting code to guarantee uniformity.
Co-authored-by: Ralf W. Grosse-Kunstleve <rwgk@google.com>
* Adding PyGILState_Check() in object_api<>::operator().
* Enabling PyGILState_Check() for Python >= 3.6 only.
Possibly, this explains why PyGILState_Check() cannot safely be used with Python 3.4 and 3.5:
https://github.com/python/cpython/pull/10267#issuecomment-434881587
* Adding simple micro benchmark.
* Reducing test time to minimum (purely for coverage, not for accurate results).
* Fixing silly oversight.
* Minor code organization improvement in test.
* Adding example runtimes.
* Removing capsys (just run with `-k test_callback_num_times -s` and using `.format()`.
* Adding test_unique_ptr_member (for desired PyCLIF behavior).
See also: https://github.com/pybind/pybind11/issues/2583
Does not build with upstream master or
https://github.com/pybind/pybind11/pull/2047, but builds with
https://github.com/RobotLocomotion/pybind11 and almost runs:
```
Running tests in directory "/usr/local/google/home/rwgk/forked/EricCousineau-TRI/pybind11/tests":
================================================================================= test session starts =================================================================================
platform linux -- Python 3.8.5, pytest-5.4.3, py-1.9.0, pluggy-0.13.1
rootdir: /usr/local/google/home/rwgk/forked/EricCousineau-TRI/pybind11/tests, inifile: pytest.ini
collected 2 items
test_unique_ptr_member.py .F [100%]
====================================================================================== FAILURES =======================================================================================
_____________________________________________________________________________ test_pointee_and_ptr_owner ______________________________________________________________________________
def test_pointee_and_ptr_owner():
obj = m.pointee()
assert obj.get_int() == 213
m.ptr_owner(obj)
with pytest.raises(ValueError) as exc_info:
> obj.get_int()
E Failed: DID NOT RAISE <class 'ValueError'>
test_unique_ptr_member.py:17: Failed
============================================================================= 1 failed, 1 passed in 0.06s =============================================================================
```
* unique_ptr or shared_ptr return
* new test_variant_unique_shared with vptr_holder prototype
* moving prototype code to pybind11/vptr_holder.h, adding type_caster specialization to make the bindings involving unique_ptr passing compile, but load and cast implementations are missing
* disabling GitHub Actions on pull_request (for this PR)
* disabling AppVeyor (for this PR)
* TRIGGER_SEGSEV macro, annotations for GET_STACK (vptr::get), GET_INT_STACK (pointee)
* adding test_promotion_of_disowned_to_shared
* Copying tests as-is from xxx_value_ptr_xxx_holder branch.
https://github.com/rwgk/pybind11/tree/xxx_value_ptr_xxx_holder
Systematically exercising returning and passing unique_ptr<T>, shared_ptr<T>
with unique_ptr, shared_ptr holder.
Observations:
test_holder_unique_ptr:
make_unique_pointee OK
pass_unique_pointee BUILD_FAIL (as documented)
make_shared_pointee Abort free(): double free detected
pass_shared_pointee RuntimeError: Unable to load a custom holder type from a default-holder instance
test_holder_shared_ptr:
make_unique_pointee Segmentation fault (#1138)
pass_unique_pointee BUILD_FAIL (as documented)
make_shared_pointee OK
pass_shared_pointee OK
* Copying tests as-is from xxx_value_ptr_xxx_holder branch.
https://github.com/rwgk/pybind11/tree/xxx_value_ptr_xxx_holder
Systematically exercising casting between shared_ptr<base>, shared_ptr<derived>.
* Demonstration of Undefined Behavior in handling of shared_ptr holder.
Based on https://godbolt.org/z/4fdjaW by jorgbrown@ (thanks Jorg!).
* Additional demonstration of Undefined Behavior in handling of shared_ptr holder.
* fixing up-down mixup in comment
* Demonstration of Undefined Behavior in handling of polymorphic pointers.
(This demo does NOT involve smart pointers at all, unlike the otherwise similar test_smart_ptr_private_first_base.)
* minor test_private_first_base.cpp simplification (after discovering that this can be wrapped with Boost.Python, using boost::noncopyable)
* pybind11 equivalent of Boost.Python test similar to reproducer under #1333
* Snapshot of WIP, TODO: shared_ptr deleter with on/off switch
* Adding vptr_deleter.
* Adding from/as unique_ptr<T> and unique_ptr<T, D>.
* Adding from_shared_ptr. Some polishing.
* New tests/core/smart_holder_poc_test.cpp, using Catch2.
* Adding in vptr_deleter_guard_flag.
* Improved labeling of TEST_CASEs.
* Shuffling existing TEST_CASEs into systematic matrix.
* Implementing all [S]uccess tests.
* Implementing all [E]xception tests.
* Testing of exceptions not covered by the from-as matrix.
* Adding top-level comment.
* Converting from methods to factory functions (no functional change).
* Removing obsolete and very incomplete test (replaced by Catch2-based test).
* Removing stray file.
* Adding type_caster_bare_interface_demo.
* Adding shared_ptr<mpty>, shared_ptr<mpty const> casters.
* Adding unique_ptr<mpty>, unique_ptr<mpty const> casters.
* Pure copy of `class class_` implementation in pybind11.h (master commit 98f1bbb800).
* classh.h: renaming of class_ to classh + namespace; forking test_classh_wip from test_type_caster_bare_interface_demo.
* Hard-coding smart_holder into classh.
* Adding mpty::mtxt string member.
* Adding isinstance<mpty> in type_caster::load functions.
* Adding rvalue_ref, renaming const_value_ref to lvalue_ref & removing const.
* Retrieving smart_holder pointer in type_caster<mpty>::load, and using it cast_op operators.
* Factoring out smart_holder_type_caster_load.
* Retrieving smart_holder pointer in type_caster<std::shared_ptr<mpty[ const]>>::load, and using it cast_op operators.
* Improved error messaging: Cannot disown nullptr (as_unique_ptr).
* Retrieving smart_holder pointer in type_caster<std::unique_ptr<mpty[ const]>>::load, and using it cast_op operators.
* Pure `clang-format --style=file -i` change.
* Pure `clang-format --style=file -i` change, with two `clang-format off` directives.
* Fixing oversight (discovered by flake8).
* flake8 cleanup
* Systematically setting mtxt for all rtrn_mpty_* functions (preparation, the values are not actually used yet).
* static cast handle for rtrn_cptr works by simply dropping in code from type_caster_base (marked with comments).
* static cast handle for rtrn_cref works by simply dropping in code from type_caster_base (marked with comments). rtrn_mref and rtrn_mptr work via const_cast (to add const).
* static cast handle for rtrn_valu works by simply dropping in code from type_caster_base (marked with comments). rtrn_rref raises a RuntimeError, to be investigated.
* Copying type_caster_generic::cast into type_caster<mpty> as-is (preparation for handling smart pointers).
* Pure clang-format change (applied to original type_caster_generic::cast).
* Adding comment re potential use_count data race.
* static handle cast implementations for rtrn_shmp, rtrn_shcp.
* Adding MISSING comments in operator std::unique_ptr<mpty[ const]>.
* static handle cast implementations for rtrn_uqmp, rtrn_uqcp.
* Bug fix: vptr_deleter_armed_flag_ptr has to live on the heap.
See new bullet point in comment section near the top.
The variable was also renamed to reflect its function more accurately.
* Fixing bugs discovered by ASAN. The code is now ASAN, MSAN, UBSAN clean.
* Making test_type_caster_bare_interface_demo.cpp slightly more realistic, ASAN, MSAN, UBSAN clean.
* Calling deregister_instance after disowning via unique_ptr.
* Removing enable_shared_from_this stub, simplifying existing code, clang-format.
Open question, with respect to the original code:
76a160070b/include/pybind11/pybind11.h (L1510)
To me it looks like the exact situation marked as `std::shared_ptr<Good> gp1 = not_so_good.getptr();` here: https://en.cppreference.com/w/cpp/memory/enable_shared_from_this
The comment there is: `// undefined behavior (until C++17) and std::bad_weak_ptr thrown (since C++17)`
Does the existing code have UB pre C++17?
I'll leave handling of enable_shared_from_this for later, as the need arises.
* Cosmetical change around helper functions.
* Using type_caster_base<mpty>::src_and_type directly, removing copy. Also renaming one cast to cast_const_raw_ptr, for clarity.
* Fixing clang-format oversight.
* Using factored-out make_constructor (PR #2798), removing duplicate code.
* Inserting additional assert to ensure a returned unique_ptr is always a new Python instance.
* Adding minor comment (change to internals needed to distinguish uninitialized/disowned in error message).
* Factoring out find_existing_python_instance().
* Moving factored-out make_constructor to test_classh_wip.cpp, restoring previous version of cast.h. This is currently the most practical approach. See PR #2798 for background.
* Copying classh type_casters from test_classh_wip.cpp UNMODIFIED, as a baseline for generalizing the code.
* Using pybind11/detail/classh_type_casters.h from test_classh_wip.cpp.
* Adding & using PYBIND11_CLASSH_TYPE_CASTERS define.
* Adding test_classh_inheritance, currently failing (passes with class_).
* Removing .clang-format before git rebase master (where the file was added).
* Bringing back .clang-format, the previous rm was a bad idea.
* Folding in modified_type_caster_generic_load_impl, just enough to pass test_class_wip. test_classh_inheritance is still failing, but with a different error: [RuntimeError: Incompatible type (as_raw_ptr_unowned).]
* Minimal changes needed to pass test_classh_inheritance.
* First pass adjusting try_implicit_casts and try_load_foreign_module_local to capture loaded_v_h, but untested and guarded with pybind11_failure("Untested"). This was done mainly to determine general feasibility. Note the TODO in pybind11.h, where type_caster_generic::local_load is currently hard-coded. test_classh_wip and test_classh_inheritance still pass, as before.
* Decoupling generic_type from type_caster_generic.
* Changes and tests covering classh_type_casters try_implicit_casts.
* Minimal test covering classh_type_casters load_impl Case 2b.
* Removing stray isinstance<T>(src): it interferes with the py::module_local feature. Adding missing #includes.
* Tests for classh py::module_local() feature.
* Pure renaming of function names in test_classh_inheritance, similar to the systematic approach used in test_class_wip. NO functional changes.
* Pure renaming of function and variable names, for better generalization when convoluting with inheritance. NO functional changes.
* Adopting systematic naming scheme from test_classh_wip. NO functional changes.
* Moving const after type name, for functions that cover a systematic scheme. NO functional changes.
* Adding smart_holder_type_caster_load::loaded_as_shared_ptr, currently bypassing smart_holder shared_ptr tracking completely, but the tests pass and are sanitizer clean.
* Removing rtti_held from smart_holder. See updated comment.
* Cleaning up loaded_as_raw_ptr_unowned, loaded_as_shared_ptr.
* Factoring out convert_type and folding into loaded_as_unique_ptr.
* Folding convert_type into lvalue_ref and rvalue_ref paths. Some smart_holder_type_caster_load cleanup.
* Using unique_ptr in local_load to replace static variable. Also adding local_load_safety_guard.
* Converting test_unique_ptr_member to using classh: fully working, ASAN, MSAN, UBSAN clean.
* Removing debugging comments (GET_STACK, GET_INT_STACK). cast.h is identical to current master again, pybind11.h only has the generic_type::initialize(..., &type_caster_generic::local_load) change.
* Purging obsolete pybind11/vptr_holder.h and associated test.
* Moving several tests to github.com/rwgk/rwgk_tbx/tree/main/pybind11_tests
a2c2f88174
These tests are from experimenting, and for demonstrating UB in pybind11 multiple inheritance handling ("first_base"), to be fixed later.
* Adding py::smart_holder support to py::class_, purging py::classh completely.
* Renaming files in include directory, creating pybind11/smart_holder.h.
* Renaming all "classh" to "smart_holder" in pybind11/detail/smart_holder_type_casters.h.
The user-facing macro is now PYBIND11_SMART_HOLDER_TYPE_CASTERS.
* Systematically renaming tests to use "class_sh" in the name.
* Renaming test_type_caster_bare_interface_demo to test_type_caster_bare_interface.
* Renaming new tests/core subdirectory to tests/pure_cpp.
* Adding new tests to CMake config, resetting CI config.
* Changing CMake file so that test_class_sh_module_local.py actually runs.
* clang-tidy fixes.
* 32-bit compatibility.
* Reusing type_caster_base make_copy_constructor, make_move_constructor with a trick.
* CMake COMPARE NATURAL is not available with older versions.
* Adding copyright notices to new header files.
* Explicitly define copy/move constructors/assignments.
* Adding new header files to tests/extra_python_package/test_files.py.
* Adding tests/pure_cpp/CMakeLists.txt.
* Making use of the new find_existing_python_instance() function factored out with PR #2822.
* Moving define PYBIND11_SMART_HOLDER_TYPE_CASTERS(T) down in the file. NO functional changes. Preparation for follow-up work (to keep that diff smaller).
* Reintroducing py::classh, this time as a simple alias for py::class_<U, py::smart_holder>.
* Replacing detail::is_smart_holder<H> in cast.h with detail::is_smart_holder_type_caster<T>.
Moving get_local_load_function_ptr, init_instance_for_type to smart_holder_type_caster_class_hooks.
Expanding static_assert in py::type::handle_of<> to accommodate smart_holder_type_casters.
* Fixing oversight.
* Adding classu alias for class_<U, std::unique_ptr<U>>.
* Giving up on idea to use legacy init_instance only if is_base_of<type_caster_generic, type_caster<T>. There are use cases in the wild that define both a custom type_caster and class_.
* Removing test_type_caster_bare_interface, which was moved to the separate PR #2834.
* Moving up is_smart_holder_type_caster, to also use in cast_is_temporary_value_reference.
* Adding smart_holder_type_casters for unique_ptr with custom deleter. SEVERE CODE DUPLICATION. This commit is to establish a baseline for consolidating the unique_ptr code.
* Unification of unique_ptr, unique_ptr_with_deleter code in smart_holder_poc.h. Leads to more fitting error messages. Enables use of unique_ptr<T, D> smart_holder_type_casters also for unique_ptr<T>.
* Copying files as-is from branch test_unique_ptr_member (PR #2672).
* Adding comment, simplifying naming, cmake addition.
* Introducing PYBIND11_USE_SMART_HOLDER_AS_DEFAULT macro (tested only undefined; there are many errors with the macro defined).
* Removing test_type_caster_bare_interface, which was moved to the separate PR #2834.
* Fixing oversight introduced with commit 95425f13d6.
* Setting record.default_holder correctly for PYBIND11_USE_SMART_HOLDER_AS_DEFAULT.
With this test_class.cpp builds and even mostly runs, except
`test_multiple_instances_with_same_pointer`, which segfaults because it is
using a `unique_ptr` holder but `smart_holder` `type_caster`.
Also adding `static_assert`s to generate build errors for such situations,
but guarding with `#if 0` to first pivot to test_factory_constructors.cpp.
* Fixing up cast.h and smart_holder.h after rebase.
* Removing detail/smart_holder_type_casters.h in separate commit.
* Commenting out const in def_buffer(... const). With this, test_buffers builds and runs with PYBIND11_USE_SMART_HOLDER_AS_DEFAULT. Explanation why the const needs to be removed, or fix elsewhere, is still needed, but left for later.
* Adding test_class_sh_factory_constructors, reproducing test_factory_constructors failure. Using py::class_ in this commit, to be changed to py::classh for debugging.
* Removing include/pybind11/detail/smart_holder_type_casters.h from CMakeLists.txt, test_files.py (since it does not exist in this branch).
* Adding // DANGER ZONE reminders.
* Converting as many py::class_ to py::classh as possible, not breaking tests.
* Adding initimpl::construct() overloads, resulting in test_class_sh_factory_constructors feature parity for py::class_ and py::classh.
* Adding enable_if !is_smart_holder_type_caster to existing initimpl::construct(). With this test_factory_constructors.cpp builds with PYBIND11_USE_SMART_HOLDER_AS_DEFAULT.
* Disabling shared_ptr&, shared_ptr* tests when building with PYBIND11_USE_SMART_HOLDER_AS_DEFAULT for now, pending work on smart_holder_type_caster<shared_ptr>.
* Factoring out struct and class definitions into anonymous namespace. Preparation for building with PYBIND11_USE_SMART_HOLDER_AS_DEFAULT.
* Simplifying from_unique_ptr(): typename D = std::default_delete<T> is not needed. Factoring out is_std_default_delete<T>() for consistentcy between ensure_compatible_rtti_uqp_del() and from_unique_ptr().
* Introducing PYBIND11_SMART_POINTER_HOLDER_TYPE_CASTERS. Using it in test_smart_ptr.cpp. With this test_smart_ptr builds with PYBIND11_USE_SMART_HOLDER_AS_DEFAULT and all but one test run successfully.
* Introducing 1. type_caster_for_class_, used in PYBIND11_MAKE_OPAQUE, and 2. default_holder_type, used in stl_bind.h.
* Using __VA_ARGS__ in PYBIND11_SMART_POINTER_HOLDER_TYPE_CASTERS.
* Replacing condense_for_macro with much simpler approach.
* Softening static_assert, to only check specifically that smart_holder is not mixed with type_caster_base, and unique_ptr/shared_ptr holders are not mixed with smart_holder_type_casters.
* Adding PYBIND11_SMART_POINTER_HOLDER_TYPE_CASTERS in test_class.cpp (with this all but one test succeed with PYBIND11_USE_SMART_HOLDER_AS_DEFAULT).
* Adding remaining PYBIND11_SMART_POINTER_HOLDER_TYPE_CASTERS. static_assert for "necessary conditions" for both types of default holder, static_assert for "strict conditions" guarded by new PYBIND11_STRICT_ASSERTS_CLASS_HOLDER_VS_TYPE_CASTER_MIX. All tests build & run as before with unique_ptr as the default holder, all tests build for smart_holder as the default holder, even with the strict static_assert.
* Introducing check_is_smart_holder_type_caster() function for runtime check, and reinterpreting record.default_holder as "uses_unique_ptr_holder". With this test_smart_ptr succeeds. (All 42 tests build, 35 tests succeed, 5 run but have some failures, 2 segfault.)
* Bug fix: Adding have_value() to smart_holder_type_caster_load. With this test_builtin_casters succeeds. (All 42 tests build, 36 tests succeed, 5 run but have some failures, 1 segfault.)
* Adding unowned_void_ptr_from_direct_conversion to modified_type_caster_generic_load_impl. This fixes the last remaining segfault (test_numpy_dtypes). New stats for all tests combined: 12 failed, 458 passed.
* Adding "Lazy allocation for unallocated values" (for old-style __init__) into load_value_and_holder. Deferring destruction of disowned holder until clear_instance, to remain inspectable for "uninitialized" or "disowned" detection. New stats for all tests combined: 5 failed, 465 passed.
* Changing std::shared_ptr pointer/reference to const pointer/reference. New stats for all tests combined: 4 failed, 466 passed.
* Adding return_value_policy::move to permissible policies for unique_ptr returns. New stats for all tests combined: 3 failed, 467 passed.
* Overlooked flake8 fixes.
* Manipulating failing ConstructorStats test to pass, to be able to run all tests with ASAN.
This version of the code is ASAN clean with unique_ptr or smart_holder as the default.
This change needs to be reverted after adopting the existing move-only-if-refcount-is-1
logic used by type_caster_base.
* Adding copy constructor and move constructor tracking to atyp. Preparation for a follow-up change in smart_holder_type_caster, to make this test sensitive to the changing behavior.
[skip ci]
* Removing `operator T&&() &&` from smart_holder_type_caster, for compatibility with the behavior of type_caster_base. Enables reverting 2 of 3 test manipulations applied under commit 249df7cbdb. The manipulation in test_factory_constructors.py is NOT reverted in this commit.
[skip ci]
* Fixing unfortunate editing mishap. This reverts the last remaining test manipulation in commit 249df7cbdb and makes all existing unit tests pass with smart_holder as default holder.
* GitHub CI clang-tidy fixes.
* Adding messages to terse `static_assert`s, for pre-C++17 compatibility.
* Using @pytest.mark.parametrize to run each assert separately (to see all errors, not just the first).
* Systematically removing _atyp from function names, to make the test code simpler.
* Using re.match to accommodate variable number of intermediate MvCtor.
* Also removing `operator T()` from smart_holder_type_caster, to fix gcc compilation errors. The only loss is pass_rref in test_class_sh_basic.
* Systematically replacing `detail::enable_if_t<...smart_holder...>` with `typename std::enable_if<...smart_holder...>::type`. Attempt to work around MSVC 2015 issues, to be tested via GitHub CI. The idea for this change originates from this comment: https://github.com/pybind/pybind11/issues/1616#issuecomment-444536813
* Importing re before pytest after observing a PyPy CI flake when importing pytest first.
* Copying MSVC 2015 compatibility change from branch pr2672_use_smart_holder_as_default.
* Introducing is_smart_holder_type_caster_base_tag, to keep smart_holder code more disconnected.
* Working around MSVC 2015 bug.
* Expanding comment for MSVC 2015 workaround.
* Systematically changing std::enable_if back to detail::enable_if_t, effectively reverting commit 5d4b6890a3.
* Removing unused smart_holder_type_caster_load::loaded_as_rvalue_ref (it was an oversight that it was not removed with commit 23036a45eb).
* Removing py::classu, because it does not seem useful enough.
* Reverting commit 6349531306 by un-commenting `const` in `def_buffer(...)`. To make this possible, `operator T const&` and `operator T const*` in `smart_holder_type_caster` need to be marked as `const` member functions.
* Adding construct() overloads for constructing smart_holder from alias unique_ptr, shared_ptr returns.
* Adding test_class_sh_factory_constructors.cpp to tests/CMakeLists.txt (fixes oversight, this should have been added long before).
* Compatibility with old clang versions (clang 3.6, 3.7 C++11).
* Cleaning up changes to existing unit tests.
* Systematically adding SMART_HOLDER_WIP tag. Removing minor UNTESTED tags (only the throw are not actually exercised, investing time there has a high cost but very little benefit).
* Splitting out smart_holder_type_casters again, into new detail/smart_holder_type_casters_inline_include.h.
* Splitting out smart_holder_init_inline_include.h.
* Adding additional new include files to CMakeLists.txt, tests/extra_python_package/test_files.py.
* clang-format cleanup of most smart_holder code.
* Adding source code comments in response to review.
* Simple micro-benchmark ("ubench") comparing runtime performance for several holders.
Tested using github.com/rwgk/pybind11_scons and Google-internal build system.
Sorry, no cmake support at the moment.
First results: https://docs.google.com/spreadsheets/d/1InapCYws2Gt-stmFf_Bwl33eOMo3aLE_gc9adveY7RU/edit#gid=0
* Breaking out number_bucket.h, adding hook for also collecting performance data for PyCLIF.
* Accounting for ubench in MANIFEST.in (simply prune, for now).
* Smarter determination of call_repetitions.
[skip ci]
* Also scaling performance data to PyCLIF.
[skip ci]
* Adding ubench/python/number_bucket.clif here for general visibility.
* Fix after rebase
* Merging detail/smart_holder_init_inline_include.h into detail/init.h.
* Renaming detail/is_smart_holder_type_caster.h -> detail/smart_holder_sfinae_hooks_only.h.
* Renaming is_smart_holder_type_caster -> type_uses_smart_holder_type_caster for clarity.
* Renaming type_caster_type_is_smart_holder_type_caster -> wrapped_type_uses_smart_holder_type_caster for clarity.
* Renaming is_smart_holder_type_caster_base_tag -> smart_holder_type_caster_base_tag for simplicity.
* Adding copyright notices and minor colateral cleanup.
* iwyu cleanup (comprehensive only for cast.h and smart_holder*.h files).
* Fixing `git rebase master` accident.
* Moving large `pragma warning` block from pybind11.h to detail/common.h.
* Fixing another `git rebase master` accident.
* Adding move_only_holder_caster `typename SFINAE = void` to enable external specializations.
* Adding SFINAE hook also to copyable_holder_caster, for uniformity, with comment to explain the purpose.
* Always call PyNumber_Index when casting from Python to a C++ integral type, also pre-3.8
* Fixed on PyPy
* Simplify use of PyNumber_Index, following @rwgk's idea, and ignore warnings in >=3.8
* Reproduce mismatch between pre-3.8 and post-3.8 behavior on __index__ throwing TypeError
* Fix tests on 3.6 <= Python < 3.8
* No, I don't have an uninitialized variable
* Fix use of __index__ on Python 2
* Make types in test_int_convert more ~boring~ descriptive
* 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>
* Only allow integer type_caster to call __int__ or __index__ method when conversion is allowed
* Remove tests for __index__ as this seems to only be used to convert to int in 3.8+
* Take both `int` and `long` types into account for Python 2
* Add test_numpy_int_convert to assert tests currently fail, even though np.intc has an __index__ method
* Also consider __index__ as noconvert to a C++ integer
* New-style classes for Python 2.7; sigh
* Add some tests on types with custom __index__ method
* Ignore some tests in Python <3.8
* Update comment about conversion from np.float32 to C++ int
* Workaround difference between CPython and PyPy's different PyIndex_Check (unnoticed because we currently don't have PyPy >= 3.8)
* Avoid ICC segfault with py::arg()
* 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.
* Fail on passing py::object with wrong Python type to py::object subclass using PYBIND11_OBJECT macro
* Split off test_non_converting_constructors from test_constructors
* Fix test_as_type, as py::type constructor now throws an error itself if the argument is not a type
* Replace tp_name access by pybind11::detail::get_fully_qualified_tp_name
* Move forward-declaration of get_fully_qualified_tp_name to detail/common.h
* Don't add the builtins module name in get_fully_qualified_tp_name for PyPy
* Add PYBIND11_BUILTINS_MODULE macro, and use it in get_fully_qualified_tp_name
* Remove code inside 'PYPY_VERSION_NUM < 0x06000000' preprocessor if branch
* fix: more cleanup
* Remove more references to PyPy 5.7 and 5.9 in the docs
* Update comment on PyUnicode_UTF* in PyPy
Co-authored-by: Henry Schreiner <henryschreineriii@gmail.com>
* Wrap PYBIND11_OVERLOAD_NAME and PYBIND11_OVERLOAD_PURE_NAME in do { ... } while (false), and resolve trailing semicolon
* Deprecate PYBIND11_OVERLOAD_* and get_overload in favor of PYBIND11_OVERRIDE_* and get_override
* Correct erroneous usage of 'overload' instead of 'override' in the implementation and internals
* Fix tests to use non-deprecated PYBIND11_OVERRIDE_* macros
* Update docs to use override instead of overload where appropriate, and add warning about deprecated aliases
* Add semicolons to deprecated PYBIND11_OVERLOAD macros to match original behavior
* Remove deprecation of PYBIND11_OVERLOAD_* macros and get_overload
* Add note to changelog and upgrade guide
* feat: type<T>()
* refactor: using py::type as class
* refactor: py::object as base
* wip: tigher api
* refactor: fix conversion and limit API further
* docs: some added notes from @EricCousineau-TRI
* refactor: use py::type::of
* fix: support nvcc and test
* fixup! fix: support nvcc and test
* docs: mention what compilers fail
* fix: much simpler logic
* refactor: slightly faster / clearer
* Fix warning C26817: Potentially expensive copy of variable 'vh' in range-for loop. Consider making it a const reference (es.71).
* Replace another instance of `for (auto vh : values_and_holders(...))` with `auto vh &` (found by @bstaletic)
Co-authored-by: Michael Goulding <Michael.Goulding@microsoft.com>
Co-authored-by: Yannick Jadoul <yannick.jadoul@belgacom.net>
* Change NAMESPACE_BEGIN and NAMESPACE_END macros into PYBIND11_NAMESPACE_BEGIN and PYBIND11_NAMESPACE_END
* Fix sudden HomeBrew 'python not installed' error
* Sweep difference in 'Class.__init__() must be called when overriding __init__' error message between CPython and PyPy under the rug
* Homebrew updated to 3.8 yesterday.
Co-authored-by: Henry Schreiner <HenrySchreinerIII@gmail.com>
This adds support for a `py::args_kw_only()` annotation that can be
specified between `py::arg` annotations to indicate that any following
arguments are keyword-only. This allows you to write:
m.def("f", [](int a, int b) { /* ... */ },
py::arg("a"), py::args_kw_only(), py::arg("b"));
and have it work like Python 3's:
def f(a, *, b):
# ...
with respect to how `a` and `b` arguments are accepted (that is, `a` can
be positional or by keyword; `b` can only be specified by keyword).
Currently user specializations of the form
template <typename itype> struct polymorphic_type_hook<itype, std::enable_if_t<...>> { ... };
will fail if itype is also polymorphic, because the existing specialization will also
be enabled, which leads to 2 equally viable candidates. With this change, user provided
specializations have higher priority than the built in specialization for polymorphic types.
This commit introduces the use of C++17-style fold expressions when
casting tuples & the argument lists of functions.
This change can improve performance of the resulting bindings: because
fold expressions have short-circuiting semantics, pybind11 e.g. won't
try to cast the second argument of a function if the first one failed.
This is particularly effective when working with functions that have
many overloads with long argument lists.
* test pair-copyability on C++17 upwards
The stdlib falsely detects containers like M=std::map<T, U>
as copyable, even when one of T and U is not copyable.
Therefore we cannot rely on the stdlib dismissing std::pair<T, M>
by itself, even on C++17.
* fix is_copy_assignable
bind_map used std::is_copy_assignable which suffers from the same problems
as std::is_copy_constructible, therefore the same fix has been applied.
* created tests for copyability
Don't assume that just because the language version is C++17 that the
standard library offers all C++17 features, too. When using clang-6.0
and --std=c++17 on Ubuntu 18.04 with libstdc++, __cpp_sized_deallocation
is false.
Pybind11 provides a cast operator between opaque void* pointers on the
C++ side and capsules on the Python side. The py::cast<void *>
expression was not aware of this possibility and incorrectly triggered a
compile-time assertion ("Unable to cast type to reference: value is
local to type caster") that is now fixed.
* Support C++17 aligned new statement
This patch makes pybind11 aware of nonstandard alignment requirements in
bound types and passes on this information to C++17 aligned 'new'
operator. Pre-C++17, the behavior is unchanged.
* Fix potential crash when calling an overloaded function
The crash would occur if:
- dispatcher() uses two-pass logic (because the target is overloaded and some arguments support conversions)
- the first pass (with conversions disabled) doesn't find any matching overload
- the second pass does find a matching overload, but its return value can't be converted to Python
The code for formatting the error message assumed `it` still pointed to the selected overload,
but during the second-pass loop `it` was nullptr. Fix by setting `it` correctly if a second-pass
call returns a nullptr `handle`. Add a new test that segfaults without this fix.
* Make overload iteration const-correct so we don't have to iterate again on second-pass error
* Change test_error_after_conversions dependencies to local classes/variables
* stl.h: propagate return value policies to type-specific casters
Return value policies for containers like those handled in in 'stl.h'
are currently broken.
The problem is that detail::return_value_policy_override<C>::policy()
always returns 'move' when given a non-pointer/reference type, e.g.
'std::vector<...>'.
This is sensible behavior for custom types that are exposed via
'py::class_<>', but it does not make sense for types that are handled by
other type casters (STL containers, Eigen matrices, etc.).
This commit changes the behavior so that
detail::return_value_policy_override only becomes active when the type
caster derives from type_caster_generic.
Furthermore, the override logic is called recursively in STL type
casters to enable key/value-specific behavior.
The current code requires implicitly that integral types are cast-able to floating point. In case of strongly-typed integrals (e.g. as explained at http://www.ilikebigbits.com/blog/2014/5/6/type-safe-identifiers-in-c) this is not always the case.
This commit uses SFINAE to move the numeric conversions into separate `cast()` implementations to avoid the issue.
* Add basic support for tag-based static polymorphism
Sometimes it is possible to look at a C++ object and know what its dynamic type is,
even if it doesn't use C++ polymorphism, because instances of the object and its
subclasses conform to some other mechanism for being self-describing; for example,
perhaps there's an enumerated "tag" or "kind" member in the base class that's always
set to an indication of the correct type. This might be done for performance reasons,
or to permit most-derived types to be trivially copyable. One of the most widely-known
examples is in LLVM: https://llvm.org/docs/HowToSetUpLLVMStyleRTTI.html
This PR permits pybind11 to be informed of such conventions via a new specializable
detail::polymorphic_type_hook<> template, which generalizes the previous logic for
determining the runtime type of an object based on C++ RTTI. Implementors provide
a way to map from a base class object to a const std::type_info* for the dynamic
type; pybind11 then uses this to ensure that casting a Base* to Python creates a
Python object that knows it's wrapping the appropriate sort of Derived.
There are a number of restrictions with this tag-based static polymorphism support
compared to pybind11's existing support for built-in C++ polymorphism:
- there is no support for this-pointer adjustment, so only single inheritance is permitted
- there is no way to make C++ code call new Python-provided subclasses
- when binding C++ classes that redefine a method in a subclass, the .def() must be
repeated in the binding for Python to know about the update
But these are not much of an issue in practice in many cases, the impact on the
complexity of pybind11's innards is minimal and localized, and the support for
automatic downcasting improves usability a great deal.
- PYBIND11_MAKE_OPAQUE now takes ... rather than a single argument and
expands it with __VA_ARGS__; this lets templated, comma-containing
types get through correctly.
- Adds a new macro PYBIND11_TYPE() that lets you pass the type into a
macro as a single argument, such as:
PYBIND11_OVERLOAD(PYBIND11_TYPE(R<1,2>), PYBIND11_TYPE(C<3,4>), func)
Unfortunately this only works for one macro call: to forward the
argument on to the next macro call (without the processor breaking it
up again) requires also adding the PYBIND11_TYPE(...) to type macro
arguments in the PYBIND11_OVERLOAD_... macro chain.
- updated the documentation with these two changes, and use them at a couple
places in the test suite to test that they work.
The `py::args` or `py::kwargs` arguments aren't properly referenced
when added to the function_call arguments list: their reference counts
drop to zero if the first (non-converting) function call fails, which
means they might be cleaned up before the second pass call runs.
This commit adds a couple of extra `object`s to the `function_call`
where we can stash a reference to them when needed to tie their
lifetime to the function_call object's lifetime.
(Credit to YannickJadoul for catching and proposing a fix in #1223).
Pybind11's default conversion to int always produces a long on Python 2 (`int`s and `long`s were unified in Python 3). This patch fixes `int` handling to match Python 2 on Python 2; for short types (`size_t` or smaller), the number will be returned as an `int` if possible, otherwise `long`. Requires Python 2.5+.
This is needed for things like `sys.exit`, which refuse to accept a `long`.
Building with the (VS2017) /permissive- flag puts the compiler into
stricter standards-compliant mode. It shouldn't cause the compiler to
work differently--it just disallows some non-conforming code--so should
be perfectly fine for the test suite under all VS2017 builds.
This commit also fixes one failure under non-permissive mode.
This changes the caster to return a reference to a (new) local `CharT`
type caster member so that binding lvalue-reference char arguments
works (currently it results in a compilation failure).
Fixes#1116
`type_descr` is now applied only to the final signature so that it only
marks the argument types, but not nested types (e.g. for tuples) or
return types.
The current C++14 constexpr signatures don't require relaxed constexpr,
but only `auto` return type deduction. To get around this in C++11,
the type caster's `name()` static member functions are turned into
`static constexpr auto` variables.
The current PYBIND11_INTERNALS_ID depends on the version of the library
in order to isolate binary incompatible internals capsules. However,
this does not preclude conflicts between modules built from different
(binary incompatible) commits with the same version number.
For example, if one module was built with an early v2.2.dev and
submitted to PyPI, it could not be loaded alongside a v2.2.x release
module -- it would segfault because of incompatible internals with
the same ID.
This PR changes the ID to depend on PYBIND11_INTERNALS_VERSION which is
independent of the main library version. It's an integer which should be
incremented whenever a binary incompatible change is made to internals.
PYBIND11_INTERNALS_KIND is also introduced for a similar reason.
The same versioning scheme is also applied to `type_info` and the
`module_local` type attribute.
The main point of `py::module_local` is to make the C++ -> Python cast
unique so that returning/casting a C++ instance is well-defined.
Unfortunately it also makes loading unique, but this isn't particularly
desirable: when an instance contains `Type` instance there's no reason
it shouldn't be possible to pass that instance to a bound function
taking a `Type` parameter, even if that function is in another module.
This commit solves the issue by allowing foreign module (and global)
type loaders have a chance to load the value if the local module loader
fails. The implementation here does this by storing a module-local
loading function in a capsule in the python type, which we can then call
if the local (and possibly global, if the local type is masking a global
type) version doesn't work.
This allows you to use:
cls.def(py::init(&factory_function));
where `factory_function` returns a pointer, holder, or value of the
class type (or a derived type). Various compile-time checks
(static_asserts) are performed to ensure the function is valid, and
various run-time type checks where necessary.
Some other details of this feature:
- The `py::init` name doesn't conflict with the templated no-argument
`py::init<...>()`, but keeps the naming consistent: the existing
templated, no-argument one wraps constructors, the no-template,
function-argument one wraps factory functions.
- If returning a CppClass (whether by value or pointer) when an CppAlias
is required (i.e. python-side inheritance and a declared alias), a
dynamic_cast to the alias is attempted (for the pointer version); if
it fails, or if returned by value, an Alias(Class &&) constructor
is invoked. If this constructor doesn't exist, a runtime error occurs.
- for holder returns when an alias is required, we try a dynamic_cast of
the wrapped pointer to the alias to see if it is already an alias
instance; if it isn't, we raise an error.
- `py::init(class_factory, alias_factory)` is also available that takes
two factories: the first is called when an alias is not needed, the
second when it is.
- Reimplement factory instance clearing. The previous implementation
failed under python-side multiple inheritance: *each* inherited
type's factory init would clear the instance instead of only setting
its own type value. The new implementation here clears just the
relevant value pointer.
- dealloc is updated to explicitly set the leftover value pointer to
nullptr and the `holder_constructed` flag to false so that it can be
used to clear preallocated value without needing to rebuild the
instance internals data.
- Added various tests to test out new allocation/deallocation code.
- With preallocation now done lazily, init factory holders can
completely avoid the extra overhead of needing an extra
allocation/deallocation.
- Updated documentation to make factory constructors the default
advanced constructor style.
- If an `__init__` is called a second time, we have two choices: we can
throw away the first instance, replacing it with the second; or we can
ignore the second call. The latter is slightly easier, so do that.
We currently allocate instance values when creating the instance itself
(except when constructing the instance for a `cast()`), but there is no
particular reason to do so: the instance itself and the internals (for
a non-simple layout) are allocated via Python, with no reason to
expect better locality from the invoked `operator new`. Moreover, it
makes implementation of factory function constructors trickier and
slightly less efficient: they don't use the pre-eallocate the memory,
which means there is a pointless allocation and free.
This commit makes the allocation lazy: instead of preallocating when
creating the instance, the allocation happens when the instance is
first loaded (if null at that time).
In addition to making it more efficient to deal with cases that don't
need preallocation, this also allows for a very slight performance
increase by not needing to look up the instances types during
allocation. (There is a lookup during the eventual load, of course, but
that is happening already).
This adds a PYBIND11_NAMESPACE macro that expands to the `pybind11`
namespace with hidden visibility under gcc-type compilers, and otherwise
to the plain `pybind11`. This then forces hidden visibility on
everything in pybind, solving the visibility issues discussed at end
end of #949.
Attempting to mix py::module_local and non-module_local classes results
in some unexpected/undesirable behaviour:
- if a class is registered non-local by some other module, a later
attempt to register it locally fails. It doesn't need to: it is
perfectly acceptable for the local registration to simply override
the external global registration.
- going the other way (i.e. module `A` registers a type `T` locally,
then `B` registers the same type `T` globally) causes a more serious
issue: `A.T`'s constructors no longer work because the `self` argument
gets converted to a `B.T`, which then fails to resolve.
Changing the cast precedence to prefer local over global fixes this and
makes it work more consistently, regardless of module load order.
This commit adds a `py::module_local` attribute that lets you confine a
registered type to the module (more technically, the shared object) in
which it is defined, by registering it with:
py::class_<C>(m, "C", py::module_local())
This will allow the same C++ class `C` to be registered in different
modules with independent sets of class definitions. On the Python side,
two such types will be completely distinct; on the C++ side, the C++
type resolves to a different Python type in each module.
This applies `py::module_local` automatically to `stl_bind.h` bindings
when the container value type looks like something global: i.e. when it
is a converting type (for example, when binding a `std::vector<int>`),
or when it is a registered type itself bound with `py::module_local`.
This should help resolve potential future conflicts (e.g. if two
completely unrelated modules both try to bind a `std::vector<int>`.
Users can override the automatic selection by adding a
`py::module_local()` or `py::module_local(false)`.
Note that this does mildly break backwards compatibility: bound stl
containers of basic types like `std::vector<int>` cannot be bound in one
module and returned in a different module. (This can be re-enabled with
`py::module_local(false)` as described above, but with the potential for
eventual load conflicts).
The builtin exception handler currently doesn't work across modules
under clang/libc++ for builtin pybind exceptions like
`pybind11::error_already_set` or `pybind11::stop_iteration`: under
RTLD_LOCAL module loading clang considers each module's exception
classes distinct types. This then means that the base exception
translator fails to catch the exceptions and the fall through to the
generic `std::exception` handler, which completely breaks things like
`stop_iteration`: only the `stop_iteration` of the first module loaded
actually works properly; later modules raise a RuntimeError with no
message when trying to invoke their iterators.
For example, two modules defined like this exhibit the behaviour under
clang++/libc++:
z1.cpp:
#include <pybind11/pybind11.h>
#include <pybind11/stl_bind.h>
namespace py = pybind11;
PYBIND11_MODULE(z1, m) {
py::bind_vector<std::vector<long>>(m, "IntVector");
}
z2.cpp:
#include <pybind11/pybind11.h>
#include <pybind11/stl_bind.h>
namespace py = pybind11;
PYBIND11_MODULE(z2, m) {
py::bind_vector<std::vector<double>>(m, "FloatVector");
}
Python:
import z1, z2
for i in z2.FloatVector():
pass
results in:
Traceback (most recent call last):
File "zs.py", line 2, in <module>
for i in z2.FloatVector():
RuntimeError
This commit fixes the issue by adding a new exception translator each
time the internals pointer is initialized from python builtins: this
generally means the internals data was initialized by some other
module. (The extra translator(s) are skipped under libstdc++).
This commit adds a PYBIND11_UNSHARED_STATIC_LOCALS macro that forces a
function to have hidden visibility under gcc and gcc-compatible
compilers. gcc, in particular, needs this to to avoid sharing static
local variables across modules (which happens even under a RTLD_LOCAL
dlopen()!). clang doesn't appear to have this issue, but the forced
visibility on internal pybind functions certainly won't hurt it and icc.
This updates the workaround from #862 to use this rather than the
version-specific template.
The fix for #960 could result a type being registered multiple times if
its `__init__` is called multiple times. This can happen perfectly
ordinarily when python-side multiple inheritance is involved: for
example, with a diamond inheritance pattern with each intermediate
classes invoking the parent constructor.
With the change in #960, the multiple `__init__` calls meant
`register_instance` was called multiple times, but the deletion only
deleted it once. Thus, if a future instance of the same type was
allocated at the same location, pybind would pick it up as a registered
type.
This fixes the issue by tracking whether a value pointer has been
registered to avoid both double-registering it. (There's also a slight
optimization of not needing to do a registered_instances lookup when the
type is known not registered, but this is secondary).
The instance registration for offset base types fails (under macOS, with
a segfault) in the presense of virtual base types. The issue occurs
when trying to `static_cast<Base *>(derived_ptr)` when `derived_ptr` has
been allocated (via `operator new`) but not initialized.
This commit fixes the issue by moving the addition to
`registered_instances` into `init_holder` rather than immediately after
value pointer allocation.
This also renames it to `init_instance` since it does more than holder
initialization now. (I also further renamed `init_holder_helper` to
`init_holder` since `init_holder` isn't used anymore).
Fixes#959.
Pre-C++17, std::pair can technically have an copy constructor even
though it can't actually be invoked without a compilation failure (due
to the underlying types being non-copyable). Most stls, including
libc++ since ~3.4, use the C++17 behaviour of not exposing an uncallable
copy constructor, but FreeBSD deliberately broke their libc++ to
preserve the nonsensical behaviour
(https://svnweb.freebsd.org/base?view=revision&revision=261801).
This updates pybind's internal `is_copy_constructible` to also detect
the std::pair case under pre-C++17.
This also everything (except for a couple cases in the internal version)
to use the internal `is_copy_constructible` rather than
`std::is_copy_constructible`.
This adds support for implicit conversions to bool from Python types
with `__bool__` (Python 3) or `__nonzero__` (Python 2) attributes, and
adds direct (i.e. non-converting) support for numpy bools.
This changes the pointer `cast()` in `PYBIND11_TYPE_CASTER` to recognize
the `take_ownership` policy: if casting a pointer with take-ownership,
the `cast()` now recalls `cast()` with a dereferenced rvalue (rather
than the previous code, which was always calling it with a const lvalue
reference), and deletes the pointer after the chained `cast()` is
complete.
This makes code like:
m.def("f", []() { return new std::vector<int>(100, 1); },
py::return_value_policy::take_ownership);
do the expected thing by taking over ownership of the returned pointer
(which is deleted once the chained cast completes).
PR #936 broke the ability to return a pointer to a stl container (and,
likewise, to a tuple) because the added deduced type matched a
non-const pointer argument: the pointer-accepting `cast` in
PYBIND11_TYPE_CASTER had a `const type *`, which is a worse match for a
non-const pointer than the universal reference template #936 added.
This changes the provided TYPE_CASTER cast(ptr) to take the pointer by
template arg (so that it will accept either const or non-const pointer).
It has two other effects: it slightly reduces .so size (because many
type casters never actually need the pointer cast at all), and it allows
type casters to provide their untemplated pointer `cast()` that will
take precedence over the templated version provided in the macro.
The value and holder iterator code had a past-the-end iterator
dereference. While of course invalid, the dereference didn't actually
cause any problems (which is why it wasn't caught before) because the
dereferenced value is never actually used and `vector` implementations
appear to allow dereferencing the past-the-end iterator. Under a MSVC
debug build, however, it fails a debug assertion and aborts.
This amends the iterator to just store and use a pointer to the vector
(rather than adding a second past-the-end iterator member), checking the
type index against the type vector size.
