Re: [boost] [config] Rethinking feature macros?

Subject: Re: [boost] [config] Rethinking feature macros?
From: Gavin Lambert (gavinl_at_[hidden])
Date: 2017-11-06 22:07:57

On 7/11/2017 07:22, Andrey Semashev wrote:
> On 11/06/17 20:59, Peter Dimov wrote:
>> I feel like we're talking past of each other here. What is there to 
>> maintain about digitalmars.hpp? It stays the same for eternity because 
>> the compiler is no longer developed so it's impossible for it to 
>> acquire a new feature.
> 
> I'm not aware of the Digital Mars compiler activity, but if it's long 
> abandoned then we might drop that file entirely. Although I can see some 
> C++14 and 17 additions in it.

I think the point was that with purely positive feature macros, the 
digitalmars.hpp header file need never be changed, because it cannot 
acquire new features.

Whereas with negative feature macros, when C++42 adds a new feature 
someone will still have to go through every old compiler header file 
(including digitalmars.hpp) and explicitly state "this doesn't support 
that".  And code will break if someone forgets to do so.

If nothing else, it seems somewhat like a SOLID violation.

(And arguing "let's drop digitalmars" is a strawman; the same thing can 
apply to any compiler, even one of the current major ones.)