I think this is a welcome move. Looking at how numba+llvmlite makes it more
compatible,
using clang-cl would improve Pythran portability.
I agree with Gaëtan that I do not fully understand why people try so hard
to get it working with MSVC.
Ashwin
On 10/19/19 12:40 PM, Gaëtan de Menten wrote:
I personally do not care much about the compiler used, as long as the
extension works and is compatible with the rest of the ecosystem. I
wonder why it is recommended in many places that C extensions for
Windows should be compiled with MSVC, and I fear using another compiler
would cause some sort of subtle breakage...
https://docs.python.org/3/extending/windows.html
https://stackoverflow.com/questions/56210402/why-do-python-extension-modules-need-to-be-compiled-with-msvc-on-windows
But I will readily admit I am clueless on this topic...
Gaëtan
On 19/10/2019 11:35, Serge Guelton wrote:
Hi folks,
I recently managed to plug into distutils to have it use clang-cl instead of
MSVC as the default compiler for windows, which would solve all our build issue
on that platform.
So here is my question to you, Windows users: would you be happy with that
alternative?
For conda users, clang-cl is already packaged, so that should be transparent...
++
Serge
