Message242920
| Author |
lemburg |
| Recipients |
BreamoreBoy, christian.heimes, lemburg, paul.moore, steve.dower, tim.golden, zach.ware |
| Date |
2015年05月11日.21:48:09 |
| SpamBayes Score |
-1.0 |
| Marked as misclassified |
Yes |
| Message-id |
<55512395.9080004@egenix.com> |
| In-reply-to |
<1431379252.35.0.9121195899.issue23970@psf.upfronthosting.co.za> |
| Content |
On 11.05.2015 23:20, Steve Dower wrote:
>
> I guess we need a third opinion.
>
> For me, the subclasses of CCompiler are undocumented and not a guaranteed interface (people using them directly are consenting adults). They're also an eyesore, so if I can clean them up without breaking the CCompiler interface (or minor version upgrades) then I should.
The distutils interface isn't documented in all details,
so the rule of thumb by which everybody operates is that any
non-private symbol is part of the public API.
FWIW: I don't see a problem with keeping implementations
around for older MS VC versions. It's well possible that
someone might want to use them for creating a Python version
compiled with an older version of MS VC, e.g. in an
embedding scenario. And you can still have your new cleaned up
version override the default msvccompiler class. |
|