First batch of PRs

[bstaletic]

I don't feel comfortable pushing to master, so I'll post a list of what I've gone through so far.

• WIP: Python 3 enums pybind/pybind11#781 - Far from finished. Seems very useful. I tried finishing it, but got stuck. Having a "lock into py::class_ after .def" makes it hard to get done.
  • Proposed category: Major
• Functional returning a pointer policy pybind/pybind11#1043 - Something about propagation of return policies. Removes 4 lines, should be safe, but there's a possibility that tests just don't cover what those 4 lines are doing.
  • Proposed category: Fixes
• Fix casting of classes with mixed polymorphic inheritance pybind/pybind11#1084 - Bug fix. Reviewed - looks fine to merge.
  • Proposed category: Fixes
• Fix cmake cxx standard definition pybind/pybind11#1089 - Cmake C++ standard version settings. Turned into bikeshedding about cmake versions and whether cmake 3.3 is too new (in 2017). Good candidate for merge.
  • Proposed category: Minor
• Add support for CMake compiler features pybind/pybind11#1098 - Cmake compiler features, by @henryiii. I haven't given it a proper review.
  • Proposed category: To review
• feat: Slice allowing None with std::optional pybind/pybind11#1101 - Slicing with None. Simple enough, but blocked on repetition of code because cast() is not available and @henryiii claims forward declaration to be tricky.
  • Proposed category: Minor
• WIP: Adding string constructor for enum pybind/pybind11#1122 - Construct py::enum_ from std::string. A simple feature turned into a discussion about case sensitivity. I'm questioning the usefulness of the PR.
  • Proposed category: To review (bikeshed)
• Priority overload pybind/pybind11#1131 - Override overload order with py::prepend(). Jakob raised a concern about performance, but without backing it with numbers. Otherwise, no complaints and ready to be merged.
  • Proposed category: Major
• pybind11 wrapped method segfaults if a function returns unique_ptr<T> when holder is shared_ptr<T> pybind/pybind11#1138 - Inconsistent use of holder classes mess. Seems superseded by #1161. Either this or #1161 should get a more detailed review, before categorizing.
  • Proposed category: To close
• Detect and fail if using mismatched holders pybind/pybind11#1161 - Look at #1138
  • Proposed category: To review

[henryiii]

Can't you make this at least a PR? There's nothing in master. And having to copy/paste your comments in kind of defeats the purpose.

A few notes:

• For 1043: tests were added, this fixes a bug that previously was not part of the tests.
• For 1101, I think with some time and maybe some help I could get the declaration right. I've marked it WIP.
• For 1122: I think it needs to be based on whatever we do with enums. Having a string constructor for enums enables a user to turn on implicit conversion from strings, which would be very useful.
• For 1131: Jakob didn't understand the PR in the first comment, once he did (in the second comment) he didn't have any performance concerns. It was approved but never merged.

[bstaletic]

Alright, I'll make a PR and close this in a moment.

[bstaletic]

Check #2 out.