On macOS, extension library builds usually rely on the flag ``-undefined
dynamic_lookup`` to get the linker to temporarily accept missing CPython
API symbols. As the name implies, those symbols are then dynamically
resolved when Python imports such an extension library. Binaries
produced in this way are more portable since they don't contain a
hardcoded path to the CPython library. This is critical when
distributing binary wheels on PyPI, etc..
When targeting macOS>=12, XCode recently started to generate a somewhat
ominous warning:
``-undefined dynamic_lookup may not work with chained fixups``
For further detail on what chained fixups are, see the following
informative WWDC
https://developer.apple.com/videos/play/wwdc2022/110362/ (the relevant
part starts about 20mins into the video)
The message that the dynamic resolution functionality became broken.
I reported this to Apple via feedback request FB11767124.
Apple investigated the request and updated the behavior of ``ld``
starting with XCode 14.3b1+: it now disables the fixup chain linker
optimization whenever ``-undefined dynamic_lookup`` is specified. The
feedback from Apple also stated that the parameter
``-Wl,-no_fixup_chains`` should be specified on pre-14.3 XCode versions
to ensure correct behavior.
This commit realizes those changes by passing a flag to *always* disable
fixup chains. Related discussion is available in the CPython repository
(https://github.com/python/cpython/issues/97524).
Without this, it's impossible to get feature parity between detection
mechanisms. Both the pkg-config file and the cmake config set their
versions, but the python probe script didn't provide an option for this.
So you could print the compiler flags for using it, but you could not
check what you got.
```
flake8...................................................................Failed
- hook id: flake8
- exit code: 1
pybind11/setup_helpers.py:177:13: B028 No explicit stacklevel keyword argument found. The warn method from the warnings module uses a stacklevel of 1 by default. This will only show a stack trace for the line on which the warn method is called. It is therefore recommended to use a stacklevel of 2 or greater to provide more information to the user.
warnings.warn("You cannot safely change the cxx_level after setting it!")
^
```
* Use PyConfig_InitPythonConfig instead of PyConfig_InitIsolatedConfig
* add unit test for default python configuration
---------
Co-authored-by: Daniel Jacobs <daniel.jacobs@is4s.com>
* Resolve flake8 error by replacing `pytest.raises(Exception)` with `SystemError`
* Also remove the obsolete comment.
* Tweak comment instead of removing it.
* Make sure to properly untrack gc objects before freeing them
* style: pre-commit fixes
* Fix lint
* Add comment about where the original track comes from
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* docs: update changelog for v2.10.3
Signed-off-by: Henry Schreiner <henryschreineriii@gmail.com>
* chore: bump versions for 2.10.3
Signed-off-by: Henry Schreiner <henryschreineriii@gmail.com>
* chore: fix make changelog script with entry is empty
Signed-off-by: Henry Schreiner <henryschreineriii@gmail.com>
Signed-off-by: Henry Schreiner <henryschreineriii@gmail.com>
* Remove clang 10 C++20 (it broke recently), add clang 15 C++20 while we are at it.
* No luck trying clang15: Error response from daemon: manifest for silkeh/clang:15 not found: manifest unknown: manifest unknown. Giving up, this needs to be a separate PR.
* chore: update clang-tidy to 15
* Add git
* Add NOLINTNEXTLINE for assignment in if
* Update CONTRIBUTING.md
* Add NOLINTNEXTLINE where needed
* Add one more NOLINTNEXTLINE
* stl_bind: make more readable
* Another missing NOLINTNEXTLINE
* Match style elsewhere
* Apply reviewer suggestion. Mark false positive
* docs: changelog for next version
Signed-off-by: Henry Schreiner <henryschreineriii@gmail.com>
* docs: address feedback
Signed-off-by: Henry Schreiner <henryschreineriii@gmail.com>
Signed-off-by: Henry Schreiner <henryschreineriii@gmail.com>
* Revert "Systematically add `-DCMAKE_VERBOSE_MAKEFILE=ON` to obtain full command lines related to `-Wodr` (#4398)"
This reverts commit ff42f5254a.
* Set `env: VERBOSE: 1` as suggested by @henryiii
* Set `env: VERBOSE: 1` also in all other .yml files using cmake
* Create templated abstract classes KeysView, ValuesView and ItemsView, and implement them on-the-fly when wrapping any specific map type
* We don't want to wrap different ValuesView objects for double values and const double, for example, as both wrappers will be named ValuesView[float]
* Fallback to C++ names if key or values types are not wrapped
* Added a test for .keys(), .values() and .items() returning the same types for similarly-typed maps
* Fixed wrong use of auto in a declarator list: the two descriptions might have different types
* Fixes for clang-tidy issues: explicit single-argument constructor, using the 'override' keyword when overriding functions
* Bugfix for old versions of clang++, which seem to have trouble with the struct being defined inside a module, which was also needlessly ugly anyway
* Bugfix for clang++, which doesn't have some of the names in runtime uness they are specified to be static
* A fix for clang-tidy performance-inefficient-string-concatenation issues - I personally think this looks uglier, but it's probably worth it for clang-tidy to be happy
* Possible fix for clang++ linking issues - make the descriptions static constexpr to make sure they are known before linking
* Correct names for previously-wrapped types as keys/values of maps
* Bugfix - typo in type info names which caused things to segfault
* Apply suggestions from code review
Co-authored-by: Aaron Gokaslan <skylion.aaron@gmail.com>
* Use detail::remove_cvref_t instead of doing remove_cv and remove_reference separately
* Avoid names with double underscore, as they are reserved
* Improved testing for KeysView, ValuesView and ItemsView: check type names + stricter asserts
* Moved description logic to helper function in type_caster_base.h
* style: pre-commit fixes
* Fix a clang-tidy issue: do not use 'else' after 'return'
* Apply suggestion by @Skylion007, with additional trivial simplification.
Co-authored-by: Amir <aimir@local>
Co-authored-by: aimir <aimir@localhost>
Co-authored-by: Aaron Gokaslan <skylion.aaron@gmail.com>
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Ralf W. Grosse-Kunstleve <rwgk@google.com>
