Leave it for later to resolve the `ubuntu-latest bazel` failure:

* https://github.com/pybind/pybind11_bazel/issues/94

```
ERROR: /home/runner/work/pybind11_abseil/pybind11_abseil/pybind11_abseil/tests/BUILD:92:17: Linking pybind11_abseil/tests/missing_import.so failed: (Exit 1): gcc failed: error executing CppLink command (from target //pybind11_abseil/tests:missing_import.so) /usr/bin/gcc @bazel-out/k8-fastbuild/bin/pybind11_abseil/tests/missing_import.so-2.params

Use --sandbox_debug to see verbose messages from the sandbox and retain the sandbox build root for debugging
/usr/bin/ld.gold: undefs: unknown -z option
```

PiperOrigin-RevId: 659560021

This change migrates users of APIs removed in NumPy 2.0 to their recommended replacements (https://numpy.org/devdocs/numpy_2_0_migration_guide.html).

PiperOrigin-RevId: 656738213

Without this a failure of the `ubuntu-latest bazel` job cancels the `ubuntu-latest cmake` job or vice versa, which is totally not useful.

PiperOrigin-RevId: 656737649

Fixes `build_and_test / ubuntu-latest cmake` failure as reported here:

* https://github.com/pybind/pybind11_abseil/pull/25#issuecomment-2173642297

PiperOrigin-RevId: 644276088

Belated minor follow-on to https://github.com/google/pybind11k/pull/30108.

This is to avoid confusing people with an obsolete name.

Similar to: https://github.com/google/pybind11k/pull/30126

The new and unrelated `build_and_test / ubuntu-latest cmake` failure will be addressed separately (see https://github.com/pybind/pybind11_abseil/pull/25#issuecomment-2173642297).

PiperOrigin-RevId: 644009757

Compared to other string-kind `type_caster`s in pybind11, `type_caster<absl::Cord>` is unusual in that it returns `bytes`. All other string-kind `type_caster`s return `str` and can be directed to use `bytes` instead via `return_value_policy::_return_as_bytes`. There is no `return_value_policy::return_as_str` that we could use here. Introducing such a policy just for `type_caster<absl::Cord>` seems heavy-handed. The existing `return_value_policy::_clif_automatic` fits this niche case organically and is fully sufficient for the purposes of PyCLIF-pybind11.

PiperOrigin-RevId: 620916112

PiperOrigin-RevId: 617961866

It shouldn't be (and previously wasn't?!) needed, AFAIK, but
something about `use_repo_rule()` seems to require it now...

PiperOrigin-RevId: 617625049

Fixes #19.

PiperOrigin-RevId: 611461599

Fixes #19.

PiperOrigin-RevId: 611177855

PiperOrigin-RevId: 611131739

Test Python dependencies now use rules_python
Update deps

PiperOrigin-RevId: 609605937

pybind11_abseil doesn't need it; Abseil does.
(Therefore, leave it in `WORKSPACE` for now.)

PiperOrigin-RevId: 609405813

PiperOrigin-RevId: 608701939

I renamed Bazel files to have `.bazel` extensions, which matters
for pybind11_abseil due to its "floating" use of pybind11_bazel.

PiperOrigin-RevId: 608631652

PiperOrigin-RevId: 607777670

Update README for consistency with internal files. Document around issue with transitive dependency discovery for Bzlmod with http_archive.

PiperOrigin-RevId: 607701389

PiperOrigin-RevId: 607462173

Added support for Bzlmod, indicated that WORKSPACE is deprecated and will be removed at a later date. Moved Python dependency support for Bazel from virtualenv to rules_python.

PiperOrigin-RevId: 607438140

PiperOrigin-RevId: 601222599

PiperOrigin-RevId: 601195303

Following the suggestion here: https://github.com/pybind/pybind11/pull/4786#issuecomment-1885448824

Piggy-backed: Minor pre-commit auto-fix for top_level_CMakeLists.txt
PiperOrigin-RevId: 597371220

Note:

This change is the result of exploring and discarding multiple approaches to fixing the zero-refcount bug in a more general way. The only approach that worked out is this local fix in the callback code. Fundamentally, code involving `StatusOr<PyObject *>` objects is inherently unsafe and bug prone, because ownership of the Python reference is not managed automatically. Ideally use of `StatusOr<PyObject *>` would generate compilation errors (e.g. via `static_assert`), which would be easy to achieve just in the pybind11_abseil repo, but would require significant sprawling changes around the Google codebase. Unfortunately, currently this is infeasible.

The command used for manual leak checking (see `# Manual verification` comments in status_testing_no_cpp_eh_test_lib.py) was:

```
blaze run //third_party/pybind11_abseil/tests:status_testing_no_cpp_eh_test
```

The `top` command was used to visually monitor `RES` for about 10 seconds, for each test case.

PiperOrigin-RevId: 597061300

PiperOrigin-RevId: 597049635

This change builds on https://github.com/google/pywrapcc/pull/30022.

PiperOrigin-RevId: 597047935

PiperOrigin-RevId: 597043524

PiperOrigin-RevId: 597039706

Preparation for changing pybind11 handling of callbacks with `Status`, `StatusOr` returns to capture C++ exceptions.

PiperOrigin-RevId: 596700611

Secondary changes:

* pybind11 needs to be pinned to a version right before https://github.com/pybind/pybind11/pull/4786 was merged, to avoid `ctest` `permission denied` errors (to be debugged separately).

* `find_package(PythonLibs REQUIRED)` is removed. It is deprecated (https://github.com/pybind/pybind11/blob/master/docs/faq.rst) and evidently not needed.

* More complete logging.

PiperOrigin-RevId: 595435856

PiperOrigin-RevId: 590290022

PiperOrigin-RevId: 579841851

The header was meant to be deprecated for >1 year, but this was not documented.

Also adopt the

```
// IWYU pragma: always_keep // See pybind11/docs/type_caster_iwyu.rst
```

comment used in https://github.com/google/pywrapcc/pull/30073.

Currently pybind11/docs/type_caster_iwyu.rst only exists in google/pywrapcc#30073, but the intend is to upstream the changes (see PR description).

See also: https://clangd.llvm.org/design/include-cleaner

PiperOrigin-RevId: 578314470

E.g. see https://github.com/include-what-you-use/include-what-you-use/blob/master/docs/IWYUPragmas.md.

PiperOrigin-RevId: 575331556

This is a functional no-op.

Note that `_` was deprecated already in Dec 2021 (https://github.com/pybind/pybind11/pull/3423).

PiperOrigin-RevId: 575085924

It is unclear what exactly broke testing with the standard command:

```
./scripts/google_run_tests.sh --make_jobs=48
```

```
/usr/local/google/home/rwgk/tmp/pybind11_abseil/tmp_build/_deps/abseil-cpp-src/absl/strings/internal/str_format/extension.h:34:6: warning: elaborated-type-specifier for a scoped enum must not use the ‘class’ keyword
   34 | enum class FormatConversionChar : uint8_t;
      | ~~~~ ^~~~~
      |      -----
/usr/local/google/home/rwgk/tmp/pybind11_abseil/tmp_build/_deps/abseil-cpp-src/absl/strings/internal/str_format/extension.h:34:33: error: found ‘:’ in nested-name-specifier, expected ‘::’
   34 | enum class FormatConversionChar : uint8_t;
      |                                 ^
      |                                 ::
```

Updating abseil-cpp to the same version as used in the bazel WORKSPACE file fixes the cmake build failures.

PiperOrigin-RevId: 574639457

PiperOrigin-RevId: 572292285

Doc fixes: Update filename in README.md, remove obsolete and broken references to pybind11_protobuf.

Fixes in response to https://github.com/pybind/pybind11_abseil/issues/10

PiperOrigin-RevId: 568278228

PiperOrigin-RevId: 563208705

Resolves `-Wunused-parameter` warnings in external environments.

PiperOrigin-RevId: 563117864

PiperOrigin-RevId: 558257360