Shuffling code in test_smart_ptr.cpp to separate struct/class definitions from bindings code. Back-porting from smart_holder branch, to minimize diffs and potential for merge conflicts. (#2875)

Thanks Wenzel!

* [dtype]: add type() method to access type attribute of PyArray_Descr (eq. to dtype.char in Python)

* [dtype] change type() name method to char_() to be compliant with Python numpy interface

* [dtype] fix by pre-commit

* [dtype] Change comments and solutions format for test

* Clarify documentation and move note about dtype.char vs PyArray_Descr::type to a plain, non-doxygen comment

* Fix and extend tests

* Fix the supposedly fixed tests

* Fix the fixed tests again

Co-authored-by: Bertrand MICHEL <bertrand.michel@onera.fr>
Co-authored-by: Yannick Jadoul <yannick.jadoul@belgacom.net>

* Splitting out detail/type_caster_base.h from cast.h.
* Include cleanup guided by include-what-you-use 0.12 based on clang version 9.0.1-11.

* Add initial implementation

* Add few more methods

* Add tests

* Fix a typo

* Use std::string constructor which takes size

* Fix implicit sign conversion error

* Add size method and test

* Remove implicit conversion

* Fix bytearray constructors and operator std::string()

* Make implicit bytearray constructor explicit

* Rerun tests

* Add null check

* Rerun tests

* Rerun tests - 2

* Remove NULL check

Co-authored-by: OnlineCop <onlinecop@gmail.com>

Bumps [pypa/gh-action-pypi-publish](https://github.com/pypa/gh-action-pypi-publish) from v1.4.1 to v1.4.2.
- [Release notes](https://github.com/pypa/gh-action-pypi-publish/releases)
- [Commits](https://github.com/pypa/gh-action-pypi-publish/compare/v1.4.1...27b31702a0e7fc50959f5ad993c78deac1bdfc29

)

Signed-off-by: dependabot[bot] <support@github.com>

Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

On Windows, clang-cl does not understand /MP.
```
clang-cl: warning: argument unused during compilation: '/MP' [-Wunused-command-line-argument]
```
with Clang 10.0.0

* Make sure all warnings in pytest get turned into errors

* Suppress DeprecationWarnings in test_int_convert and test_numpy_int_convert

* PyLong_AsLong only shouts "Deprecated!" on Python>=3.8

* Fix remaining warnings on PyPy and CPython 3.10-dev

* chore: get PyPy 3.7 wheels using NumPy 1.20

* Add Valgrind suppressions after NumPy update

Co-authored-by: Yannick Jadoul <yannick.jadoul@belgacom.net>

* Demonstrate issue with weakref constructor overloads

* Fix weakref constructor to convert on being passed a non-weakref object

* Improve on nonlocal-scoped variable in test_weakref

* Keep backwards-compatibility by introducing PYBIND11_OBJECT_CVT_DEFAULT macro

* Simplify test_weakref

Adding holder_caster `typename SFINAE = void` hooks to help work around the current lack of smart-pointer interoperability (#2833)

* 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.

* When determining if a shared_ptr already exists, use a test on the weak_ptr instead of a try/catch block.

* When determining if a shared_ptr already exists, use a test on the weak_ptr instead of a try/catch block.

* weak_from_this is only available in C++17 and later

* Switch to use feature flag instead of C++ version flag.

* Add Microsoft-specific check.

* Avoid undefined preprocessor macro warning treated as error.

* Simplify shared_from_this in init_holder

* Include <version> in detail/common.h (~stolen~ borrowed from @bstaletic's #2816)

* Move class_::get_shared_from_this to detail::try_get_shared_from_this

* Simplify try_get_shared_from_this by using weak_ptr::lock()

Co-authored-by: Yannick Jadoul <yannick.jadoul@belgacom.net>

* Changing pybind11::str to exclusively hold PyUnicodeObject

Factoring out find_registered_python_instance() from type_caster_generic::cast.

* docs: prepare for 2.6.2

* chore: pre-commit autoupdate

Bumps [jwlawson/actions-setup-cmake](https://github.com/jwlawson/actions-setup-cmake) from v1.6 to v1.7.
- [Release notes](https://github.com/jwlawson/actions-setup-cmake/releases)
- [Commits](https://github.com/jwlawson/actions-setup-cmake/compare/v1.6...821902fd0e0da89b13fd3161d358e96bf7550883

)

Signed-off-by: dependabot[bot] <support@github.com>

Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* 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

* Force the builtin module key to be the correct type.

Previously it was always going to be a std::string which converted into
unicode. Python 2 appears to want module keys to be normal str types, so
this was breaking code that expected plain string types in the
builtins.keys() data structure

* Add a simple unit test to ensure all built-in keys are str

* Update the unit test so it will also run on pypy

* Run pre-commit.

Co-authored-by: Jesse Clemens <jesse.clemens@sony.com>

* style: adding clang-format as manual hook

* docs: adding a Clang-Format section. (#2803)

Co-authored-by: Ralf W. Grosse-Kunstleve <rwgk@google.com>

* style: avoid using unintialized variables

Tested with cmake --warn-unintialized -S . -B build

* refactor: use function for possibly uninit vars

This allows users to set e.g. `extra_compile_args=["-g"]` (temporarily,
for debugging purposes) and not have that get overridden by
Pybind11Extension's default `-g0`.

In order to minimize "order inversion" (not that it really matters,
though) I chose to collect all flags and add them at once (hence the
dropping of `*args`).  I also dropped flag deduplication, which seems
unneeded.

* 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:...

fix: only allow integer type_caster to call __int__ method when conversion is allowed; always call __index__ (#2698)

* 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()

* docs: update build description slightly

* docs: missing ticks