On the proposed ccompiler.has_flag method

[SylvainCorlay]

There is an ongoing discussion that was triggered by a proposed patch to distutils.

• The patch consists in adding a new has_flag function to distutils' ccompiler, aside of has_function to detect whether the compiler supports certain compilation flag. It is similar to cmake's CHECK_CXX_COMPILER_FLAG.

  Detecting whether a compiler support such a flag is crucial for extension authors who need e.g. C++11 support in their code, so as to be able to output meaningful error messages when not available, or to use such a flag optionally when it improves performance. This becomes crucial since we have new flavors of the C++ programming language every 3 years (C++11/14/17).

• This is obviously a backward compatible change since it is a new method, although it triggered a discussion on whether one should allow incremental improvements to distutils.

  Here are links to the top-most messages of two threads:

  1. About the proposed feature: https://mail.python.org/pipermail/distutils-sig/2016-August/029469.html
  2. What is the official position on distutils: https://mail.python.org/pipermail/distutils-sig/2016-August/029595.html

My proposal is the following:

• add the patch to distutils.ccompiler. Incremental improvement is good! More reasons to adopt more recent python versions.
• optionally add it to ccompiler via setuptools for earlier versions of python.

ping @jaraco

[jaraco]

I have no objections, and my opinion is that if someone is willing to propose the changes via a pull request and address the ramifications of those changes to stability, it will be accepted.