Sylvester
Sylvester
I have seen somewhere, that there may be Windows/Numpy 2.0 specific problem in doctest. I can check tomorrow.
It is more complicated. Change to NPY_INT64/LONGLONG is actually the reason of the last 4 failures on Windows (test passes on macOS) Numpy version doesn't matter! The simplest way to...
The only combination working seems to be: NPY_LONG def pre_process_args: dtype=np.int32 (3 occurrences, flag and ext_mask) For some reason only default c int works correctly (32 bit on Windows, 64...
After more testing it looks like only 32 integers are accepted on Windows in these arrays if compiled with MSVC and problem depends on change from 32 to 64 bit...
Removed last post - (doesn't work with 1.26) In summary: NPY_LONG has always been 32 bit on Windows, 64 bit on Linux (I am sorry for wrong info) numpy.long has...
My final thought (tested with NPY_LONG): def pre_process_args: d_type=np.long if np.__version__.startswith("2") else int
Replacing "long" with "long long" across .cpp and .h files is also working
CIBW is actually working with impressive level of automation! name: Build wheels on: workflow_dispatch: pull_request: push: branches: - master jobs: build_wheels: name: Build wheels on ${{ matrix.os }}-${{matrix.version}} runs-on: ${{...
Started my fork as private, therefore PR is not possible.
Would it be possible via e-mail: [email protected] ? I am at CET