[Development] Raising the minimum to C++20

[Allan Sandfeld Jensen]

On Mittwoch, 10. Mai 2023 08:01:11 CEST Marc Mutz via Development wrote:
> On 10.05.23 01:21, Thiago Macieira wrote:
> > On Tuesday, 2 May 2023 17:39:01 PDT Thiago Macieira wrote:
> >> Opinions?
> > 
> > BTW, here's the opinion of the GCC devs:
> > 
> > Don't ever use the -std= option to raise the language from the default.
> > That implies opting in to functionality that they're not entirely
> > satisfied with, and may possibly still break ABI.
> 
> "The GCC dev_s_", or "one GCC dev"? And who? Jonathan? [citation needed]
> 
> BTW: Did any GCC version ever default to C++11? IIRC, they went from
> 98/03 straight to 14 (which I agree with, but that's another story).
> 
Yeah. The discussion was on what was the expected default. When they suggested 
switching to C++11, I proposed just going with C++14 directly being a "bug-
fix" of C++11, and that side won out. In any case the gcc default is not just 
what they satisfied with. It is based on lazily updating stable defaults, and 
discussions on the mailing list.

Though the documentation still marks C++20 support as "experimental". So I 
guess they could justify breaking it if they wanted to. Especially the 
coroutines and modules which are not enabled by default with std=c++20. I find 
it highly unlikely they would break what works by default with std=c++20 
though, even in ABI. It is not the sort of thing that goes over well.

'Allan