* Start pybind11v3
* Remove all code for PYBIND11_INTERNALS_VERSION 4 and 5
* Fix oversight: pybind11/_version.py needs to be updated as well.
* Add @pytest.mark.skipif("env.GRAALPY", reason="TODO debug segfault")
* Backport of https://github.com/google/pybind11clif/pull/30088 (main PR) and https://github.com/google/pybind11clif/pull/30092 (minor fixes).
Note for completeness:
These are identical to the current versions on the pybind11clif main branch (@ commit 4841661df5daf26ecdedaace54e64d0782e63f64):
* test_class_release_gil_before_calling_cpp_dtor.cpp
* test_class_release_gil_before_calling_cpp_dtor.py
* Fix potential data race in test_class_release_gil_before_calling_cpp_dtor.cpp
The original intent was to let the singleton leak, but making that tread-safe is slightly more involved than this solution. It's totally fine in this case if the RegistryType destructor runs on process teardown.
See also: https://github.com/pybind/pybind11/pull/5522#issuecomment-2661068351
---------
Co-authored-by: Ralf W. Grosse-Kunstleve <rwgk@google.com>
* Remove some maybe-uninitialized warnings
In the Eigen matrix type_caster, resize the matrix instead of assigning with a new one when the matrix size needs to be adjusted.
This can remove lots of compiling warnings about "maybe-uninitialized".
* Revert "Remove some maybe-uninitialized warnings"
This reverts commit 7d5a9b41aa.
* Suppress `-Wmaybe-uninitialized` warning
Reproducer: https://github.com/pybind/pybind11/pull/5516#issuecomment-2645846295
---------
Co-authored-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
* fix docs build command on different global and venv python versions
* feat: add --extension-suffix
Signed-off-by: Henry Schreiner <henryschreineriii@gmail.com>
---------
Signed-off-by: Henry Schreiner <henryschreineriii@gmail.com>
Co-authored-by: Nikul Patel <nikul@z830>
Co-authored-by: Henry Schreiner <henryschreineriii@gmail.com>
* Added rework of arg/return typing
* Changed `Path` to `pathlib.Path` for compatibility with pybind11-stubgen
* Removed old arg/return type hint implementation
* Added noconvert support for arg/return type hints
* Added commented failing tests for Literals with special characters
* Added return_descr/arg_descr for correct typing in typing::Callable
* Fixed clang-tidy issues
* Changed io_name to have explicit return type (for C++11 support)
* style: pre-commit fixes
* Added support for nested callables
* Fixed missing include
* Fixed is_return_value constructor call
* Fixed clang-tidy issue
* Uncommented test cases for special characters in literals
* Moved literal tests to correct test case
* Added escaping of special characters in typing::Literal
* Readded mistakenly deleted bracket
* Moved sanitize_string_literal to correct namespace
* Added test for Literal with `!` and changed StringLiteral template param name
* Added test for Literal with multiple and repeated special chars
* Simplified string literal sanitization function
* Added test for `->` in literal
* Added test for `->` with io_name
* Removed unused parameter name to prevent warning
* Added escaping of `-` in literal to prevent processing of `->`
* Fixed wrong computation of sanitized string literal length
* Added cast to prevent error with MSVC
* Simplified special character check
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* Fix data race when using shared variables (free threading)
In the free threading build, there's a race between wrapper re-use and
wrapper deallocation. This can happen with a static variable accessed by
multiple threads.
Fixing this requires using some private CPython APIs: _Py_TryIncref and
_PyObject_SetMaybeWeakref. The implementations of these functions are
included until they're made available as public ("unstable") APIs.
Fixes#5489
* style: pre-commit fixes
* Avoid unused parameter
* Add missing return for default build
* Changes from review
* Assign result to local variable
* s/clang-tidy/ruff
* clang-tidy: static is redundant
* Use 'noqa: B018'
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* Fix module type hint
"module" is not a valid python value.
The correct type hint for a module object is "types.ModuleType" which has existed since at least Python 2.6
* Added module type hint test
* style: pre-commit fixes
* Remove doc function
* Fixed type hint
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* THIS IS JUST A START: First attempt to combine information from PR #4953 and PR #5437
* Include GXX_ABI and USE_CXX in the identifier
Further constrain to GXX_ABI 1002 or greater and less than 2000,
hopefully future proof by summarizing that as `1` along with CXX11 on or
off.
* style: pre-commit fixes
* Use `gxx_abi_1xxx` and simplify the Clang string
After discussions with Ralf Grosse-Kunstleve we think these would make
good identifiers that are concise and clear.
* Error if `_GLIBCXX_USE_CXX11_ABI` is not defined
Within the `__GXX_ABI_VERSION` block this should always be defined,
guard against unexpected defines and make the error obvious.
* Change `usecxx11` to `use_cxx11_abi` for correspondence with `_GLIBCXX_USE_CXX11_ABI` (similarly to `gxx_abi` for `__GXX_ABI_VERSION`).
* `PYBIND11_COMPILER_TYPE` overhaul, mainly: replace `_icc`, `_clang`, `_gcc` with `_system`
* Add NVHPC (__PGI) to the list of compilers compatible with system compiler.
See comment by @robertmaynard: https://github.com/pybind/pybind11/pull/5439#issuecomment-2498839010
* Fix oversight: remove __NVCOMPILER elif branch in PYBIND11_BUILD_ABI block.
Also add comment pointing to this PR (#5439).
* Revert "Fix oversight: remove __NVCOMPILER elif branch in PYBIND11_BUILD_ABI block."
This reverts commit d412303e72.
* Revert "Add NVHPC (__PGI) to the list of compilers compatible with system compiler."
This reverts commit 9fc9515885.
* Define NVHPC PYBIND11_BUILD_ABI using __GNUC__, __GNUC_MINOR__, _GLIBCXX_USE_CXX11_ABI
* Use _GLIBCXX_USE_CXX11_ABI to detect libstdc++, then assume that NVHPC is always in the 1xxx ABI family.
* Enhance NVHPC comment and limited future proofing.
* The `PYBIND11_STDLIB` is obsolete but kept around to maintain backward compatibility.
* Move `PYBIND11_BUILD_TYPE` down in the file, so that the order of macro definitions is the same as in the list defining `PYBIND11_PLATFORM_ABI_ID`
* Introduce `PYBIND11_COMPILER_TYPE_LEADING_UNDERSCORE`:
This makes it possible to achieve these two goals:
* Avoid the leading underscore in `PYBIND11_PLATFORM_ABI_ID` (see https://github.com/pybind/pybind11/pull/5439#issuecomment-2503762161)
* Maintain backward compatibility for use cases as reported under https://github.com/pybind/pybind11/pull/5439#issuecomment-2510212677
`PYBIND11_INTERNALS_KIND` is removed in this commit to ensure that `PYBIND11_COMPILER_TYPE` is the first element of the `PYBIND11_PLATFORM_ABI_ID`, so that `PYBIND11_COMPILER_TYPE_LEADING_UNDERSCORE` can meaningfully be used as a prefix for `PYBIND11_PLATFORM_ABI_ID` in pybind11/detail/internals.h.
* Apply suggestion by @isuruf, with revised comments (code is as suggested).
* Make determination of `PYBIND11_COMPILER_TYPE` `"macos"` or `"glibc"` more general.
The main motivation is to resolve these "Manylinux on 🐍 3.13t • GIL" and "Pyodide wheel" failures:
```
/__w/pybind11/pybind11/include/pybind11/conduit/pybind11_platform_abi_id.h:35:10: error: #error "Unknown PYBIND11_COMPILER_TYPE: PLEASE REVISE THIS CODE."
35 | # error "Unknown PYBIND11_COMPILER_TYPE: PLEASE REVISE THIS CODE."
| ^~~~~
```
(All other CI jobs succeeded.)
Further thought:
Essentially, under Linux and macOS the `PYBIND11_COMPILER_TYPE` is only for informational purposes.
ABI compatibility is determined by the libstdc++ or libc++ ABI version.
* Add `PYBIND11_COMPILER_TYPE` `emscripten`
* Add `PYBIND11_COMPILER_TYPE` `graalvm`
* Revert "Add `PYBIND11_COMPILER_TYPE` `graalvm`"
This reverts commit 75da5fbfd9.
* Revert "Add `PYBIND11_COMPILER_TYPE` `emscripten`"
This reverts commit e34dc8b511.
* Revert "Make determination of `PYBIND11_COMPILER_TYPE` `"macos"` or `"glibc"` more general."
This reverts commit 41daaa41fa.
* Revert "Apply suggestion by @isuruf, with revised comments (code is as suggested)."
This reverts commit ca9e6990de.
* Remove `defined(__INTEL_COMPILER)` as suggested by @hpkfft under https://github.com/pybind/pybind11/pull/5439#discussion_r1889156543
---------
Co-authored-by: Marcus D. Hanwell <marcus@cryos.net>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* Added arg/return type handling.
* Added support for nested arg/return type in py::typing::List
* Added support for arg/return type in stl/filesystem
* Added tests for arg/return type in stl/filesystem and py::typing::List
* Added arg/return name to more py::typing classes
* Added arg/return type to Callable[...]
* Added tests for typing container classes (also nested)
* Changed typing classes to avoid using C++14 auto return type deduction.
* Fixed clang-tidy errors.
* Changed Enable to SFINAE
* Added test for Tuple[T, ...]
* Added RealNumber with custom caster for testing typing classes.
* Added tests for Set, Iterable, Iterator, Union, and Optional
* Added tests for Callable
* Fixed Callable with ellipsis test
* Changed TypeGuard/TypeIs to use return type (being the narrower type) + Tests
* Added test for use of fallback type name with stl vector
* Updated documentation.
* Fixed unnecessary constructor call in test.
* Fixed reference counting in example type caster.
* Fixed clang-tidy issues.
* Fix for clang-tidy
* Updated cast method to use pybind11 API rather than Python C API in custom caster example
* Updated load to use pybind11 API rather than Python C API in custom caster example
* Changed test of arg/return name to use pybind11 API instead of Python C API
* Updated code in adcanced/cast example and improved documentation text
* Fixed references in custom type caster docs
* Fixed wrong logical and operator in test
* Fixed wrong logical operator in doc example
* Added comment to test about `float` vs `float | int`
* Updated std::filesystem::path docs in cast/overview section
* Remove one stray dot.
---------
Co-authored-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
* fix: make PYBIND11_WARNING_POP actually pop clang diagnostics
* fix: ignore -Wgnu-zero-variadic-macro-arguments on clang
* Revert "fix: ignore -Wgnu-zero-variadic-macro-arguments on clang"
This reverts commit d310959bf5.
* C++20 modernization: Use `__VA_OPT__(, ) __VA_ARGS__` in `PYBIND11_DECLARE_HOLDER_TYPE()`
* Disable `__VA_OPT__(, ) __VA_ARGS__` usage for MSVC (it is unclear to rwgk why it does not work).
This is the beginning of the error message:
```
D:\a\pybind11\pybind11\tests\test_smart_ptr.cpp(285,1): error C2143: syntax error: missing ')' before ',' [D:\a\pybind11\pybind11\build\tests\pybind11_tests.vcxproj]
D:\a\pybind11\pybind11\tests\test_smart_ptr.cpp(285,1): error C2059: syntax error: ')' [D:\a\pybind11\pybind11\build\tests\pybind11_tests.vcxproj]
```
* Add `PYBIND11_WARNING_DISABLE_CLANG("-Wgnu-zero-variadic-macro-arguments")` in test_smart_ptr.cpp
This is the error message:
```
/__w/pybind11/pybind11/tests/test_smart_ptr.cpp:287:51: error: must specify at least one argument for '...' parameter of variadic macro [-Werror,-Wgnu-zero-variadic-macro-arguments]
PYBIND11_DECLARE_HOLDER_TYPE(T, std::shared_ptr<T>)
^
/__w/pybind11/pybind11/include/pybind11/cast.h:885:13: note: macro 'PYBIND11_DECLARE_HOLDER_TYPE' defined here
^
```
* Also add `PYBIND11_WARNING_DISABLE_CLANG("-Wgnu-zero-variadic-macro-arguments")` in test_virtual_functions.cpp
* Also add `PYBIND11_WARNING_DISABLE_CLANG("-Wgnu-zero-variadic-macro-arguments")` in test_embed/test_interpreter.cpp
* Undo all changes except the original push -> pop fix.
* 1. Add `PYBIND11_WARNING_DISABLE_CLANG("-Wgnu-zero-variadic-macro-arguments")` near the top of pybind11/pybind11.h; 2. Change `PYBIND11_DECLARE_HOLDER_TYPE` macro to side-step the only remaining clang warning-as-error (this is still needed even for clang 18).
Alternatively the warning suppression could be moved into pybind11/cast.h, but this commit limits the warning suppression to smaller scope within include/pybind11.
---------
Co-authored-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
* Add support for array_t<handle> and array_t<object>
* style: pre-commit fixes
* Remove loops that aren't strictly needed
* Fix compiler warning
* Disable GC-dependent checks when running on pypy or graalpy
* style: pre-commit fixes
* Remove PyValueHolder counter again
* Move tests to templates to avoid code duplication
* Rerun pre-commit
* Restore import that was erroneously removed by pre-commit
* Reduce code duplication with more template magic
* Bring back `.attr("value")` in `return_array_cpp_loop()`
This was meant to further stress-test correctness of refcount handling.
All modified test functions were manually leak-checked (`while True`, top command, Python 3.12.3, Ubuntu 24.01, gcc 13.2.0).
---------
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
* feat: allow annotate methods with `pos_only` when only have the `self` argument
* chore(typing): make arguments for auto-generated dunder methods positional-only
* docs: add more comments to improve readability
* style: fix nit suggestions
* Add test_self_only_pos_only() in tests/test_methods_and_attributes
* test: add docstring tests for generated dunder methods
* test: remove failed tests
* fix(test): run `gc.collect()` three times for refcount tests
---------
Co-authored-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
* Allow subclasses of args and kwargs
The current implementation disallows subclasses of args and kwargs
* Added object type hint to args and kwargs
* Added type hinted args and kwargs classes
* Changed default type hint to typing.Any
* Removed args and kwargs type hint
* Updated tests
Modified the tests from #5381 to use the real Args and KWArgs classes
* Added comment
