Fix up the NumPy warnings hack by using the actual "import_array()" function declaration instead of a wrong one.

Also mark the case of an uninitialised memory view slice as "unlikely()" when refcounting it since it's most probably not a performance critical case.

Mark the last deletion of a memory view slice as "unlikely()" to ask the C compiler to optimise for the repeated re-use case (usually in loops).

Mark the first creation of a memory view slice as "unlikely()" to ask the C compiler to optimise for the repeated re-use case (usually in loops).

Closes https://github.com/cython/cython/issues/3594

Closes https://github.com/cython/cython/issues/1462

Move the work directory of the examples tests into an 'examples/' subdirectory to get it out of the way of the other test types.

Move the work directory of the examples tests into an 'examples/' subdirectory to get it out of the way of the other test types.

Make sure the module init function is declared as 'extern "C"' in C++ also when the CYTHON_NO_PYINIT_EXPORT option is used.
Closes #3414.

Closes #3038.

Closes #3038.

https://github.com/numpy/numpy/pull/16170/commits/e00d94b70880b1321eff727ac7eb897be7422340

Avoid MSVC warning in refnanny about an integer assignment that reduces the value range (although it's probably safe to assume that C code line numbers > 2^30 will remain rare for another while).

Avoid an additional utility code section for 2-exception matching by relying on existing special-cased macros.

For Py3.5+, where inittab should really be used instead. For GCC, Clang, MSVC and C++14