#1393 - Bug repro, not an actual PR. Talks about "overload ambiguity between Numpy arrays and scalar".
- Category: Bug reproducer
#1266 - Enables readonly memory views. Two lines change, backwards compatible. No tests. Close in favour of #1466
- Category: Close
#1301 - Bug repro, not an actual PR. Talks about a memory leak.
- Category: Bug reproducer
#1368 - Fixes a segfault with iostream wrappers and releasing GIL.
- Category: Bugfix
- @henryiii should explain the use case first.
#1373 - Cmake bugfix. Sets PythonLibsNew_FOUND in tools/FindPythonLibsNew.cmake
- Category: Bugfix
#1209 - An attempt to fix a core dump when a C++ thread running a python script gets terminated. Uses some GCC specific black magic, so I don't know what it is about. Tests are needed.
- Category: WIP
#1196 - Keyword-only arguments support.
- Category: Major
#1210 - Moves optional and variant casters into their own header. Backwards compatible. Bikeshedding made it go stale. Should be merged as is and the bikeshedding discussion, that could be proven useful, taken to a separate issue.
- Category: Minor
#1220 - Correctly document that pybind requires cmake 3.0
- Category: WIP
#1243 - Nicer error instead of std::bad_alloc in one specific case. Lacks tests.
- Category: WIP
#1255 - Bazel "strict" mode fixes.
- Category: Bugfix
#1257 - Bug repro, not an actual PR. Talks about a specific combination of virtual inheritance and diamond pattern causing a crash on MSVC.
- Category: Bug reproducer
#1260 - R-value references support for py::cast()
- Category: Minor
#1286 - Pybind GIL API is incompatible with the raw CPython API because they keep track of python state separately. This is a long and contrieved way of making them play nice. My opinion: close as it is a strange way to make these cooperate. Better solution for those that need CPython API is to have a basic pybind class without the "advance GIL use case".
- Category: Close
#1322 - Look at #1286
- Category: Close
#1325 - Allows custom numpy ufuncs to be defined. Work in progress. Stale.
- Category: WIP
#1334 - Bugfix for __int__ and __hash__ when py::enum_ has underlying type of char.
- Category: Bugfix
#1329 - Bugfix for fixed sized numpy dtypes.
- Category: Bugfix
#1192 - Allows std::enable_shared_from_this to be used for casting and lets pybind avoid a copy in some cases.
- Category: Minor
#1394 - Signed slices
- Category: Minor
#1413 - Allow builtins to be used as callbacks
- Category: Minor
#1428 - "Generalize cmake support" was greenlighted by @jakob, but was missing the update to the docs.
- Category: WIP
#1447 - WOrkaround for PGI compiler bug. Bug is fixed on PGI version 18.7. Latest version is 18.10.
- Category: Close
#1463 - py::isinstance<py::str> is too permissive. BUgfix is a WIP.
- Category: WIP
#1466 - Similar idea to memory views, but with a greater scope. Adds readonly flag to the buffer protocol. The py::array and others (including memory views) would be separate PRs.
- Category: Minor/Major
#1494 - Dynamic attribute support on Pypy 6, breaks the tests and doesn't seem to work. The only thing indicating that it works is the PR author's word.
- Category: Close
#1496 - py::bind_vector can only be extended by another object of the same type, so string_vector.extend(['asd']) doesn't work, but string_vector.extend(StringVector(['asd'])) does. One of the official maintainers called this "an overlooked feature".
- Category: Minor
#1641 - Getters for exception type value and traceback.
- Category: Minor
#1635 - Typos in documntation
- Category: Minor
#1633 - Wrong.
- Category: Close
#1629 - Another typo.
- Category: Minor