Re: USE_LANGUAGES and wrappers

[Joerg Sonnenberger <joerg%bec.de@localhost>]

On Mon, Jul 24, 2023 at 09:10:59AM -0400, Greg Troxel wrote:
> Joerg Sonnenberger <joerg%bec.de@localhost> writes:
> 
> > On Sun, Jul 23, 2023 at 07:14:54AM -0400, Greg Troxel wrote:
> >> Greg Troxel <gdt%lexort.com@localhost> writes:
> >> 
> >> >   3) Adjust wrappers so that if e.g. c++17 is not in USE_CXX_FEATURES,
> >> >   then any invocation with --std=c++17 will fail.
> >> 
> >> I believe there have been no objections to this point either.
> >
> > This is no different really than defaulting to some archaic old standard
> > for the sake of it.
> 
> It causes a failrue when a language variant should have been declared
> but wasn't.  Fixing that means that when building on a platform with a
> base compiler that doens't support that standard, an adequate compiler
> will be forced.

So when you use an outdated compiler, you will get a failure. That's no
different from lots of other situations. In that case, the tag can be
added and it didn't disrupt people on a modern system.

> Regardless of whether we force an failure, it is an error for a package
> to use c++17 if c++17 is not in USE_CXX_FEATURES.

Please don't optimize the system for outdated compilers. That's just
silly.

> Pleaes refrain from using words like archaic incorrectly.  That probably
> applies fairly to K&R C, and might apply to c89 and c++03.  But c++11 is
> merely "not up to date with rapid specification changes".

I'm fairly certain the use of the word is correct. I certainly consider
a baseline of a 6 year old compiler as outdated.

Joerg