patch to support PowerPC and Windows 32 bits

[lrineau]

Summary of Changes

Patch required for the acceptance of CGAL-5.6 in vcpkg. The CI system tests on x86_windows (32 bits), and there was an error:

D:\installed\x86-windows\include\CGAL/cpp_float.h(30): error C3861: '_BitScanForward64': identifier not found
D:\installed\x86-windows\include\CGAL/cpp_float.h(42): error C3861: '_BitScanReverse64': identifier not found

See microsoft/vcpkg#32896 (comment)

The solution is do disable support for Boost MP on Windows 32 bits. That patch has been tested with the vcpkg CI, and it worked.

Release Management

• Affected package(s): Number_types
• Issue(s) solved (if any): patch required by [cgal] update to 5.6 microsoft/vcpkg#32896

[sloriot]

Shouldn't you patch Number_types/CGAL/include/boost_mp.h instead ?

Currently the condition is

    (!defined _MSC_VER || BOOST_VERSION >= 107000)
#define CGAL_USE_BOOST_MP 1

Maybe you can use _WIN32 here?

[lrineau]

There is also a similar issue with the Conda CI, on a ppc64le platform:
https://dev.azure.com/conda-forge/feedstock-builds/_build/results?buildId=752930&view=logs&j=8c139a40-380b-5518-bf6b-edcb18cbf319&t=e71f8d0d-13be-528e-b5e7-bfc20e0db699&l=4831

[lrineau]

Shouldn't you patch Number_types/CGAL/include/boost_mp.h instead ?

Maybe. That would fix also the bug for users not using our CMake scripts.

I was more comfortable with a patch in CMake.

Currently the condition is

    (!defined _MSC_VER || BOOST_VERSION >= 107000)
#define CGAL_USE_BOOST_MP 1

Maybe you can use _WIN32 here?

_WIN32 is not a right choice because it does not really discriminate 32/64 bits:

https://learn.microsoft.com/en-us/cpp/preprocessor/predefined-macros?

_WIN32 Defined as 1 when the compilation target is 32-bit ARM, 64-bit ARM, x86, or x64. Otherwise, undefined.

Maybe !_WIN64 instead.

[lrineau]

I have a new patch. Please review cf0c6c0.

[lrineau]

@mglisse
I think we could support Windows 32 bits using that kind of code:

#ifdef _WIN64
  _BitScanForward64(&r, x);
#else
  if (_BitScanForward(&r, (uint32_t)(x >> 32))) {
    r += 32;
  } else {
    _BitScanForward(&r, (uint32_t)x);
  }
#endif

See https://github.com/janestreet/base/blob/02cf612470d5d37e0e9a704cdcecd5306ff196b7/src/int_math_stubs.c#L40-L52

Would the rest of CGAL/cpp_float.h work?

[sloriot]

Successfully tested in CGAL-6.0-Ic-39