On Fri, 30 Apr 2021 at 11:08, Jₑₙₛ Gustedt <jens.gustedt_at_[hidden]> wrote:
> > Right, attempts to introduce VLAs into C++ have failed.
>
> I don't know what is right about this. They "have not failed" in a
> passive way, but people have actively decided to make them fail.

I don't know what you mean by that. We had multiple different approaches
a couple of years ago tried, but they failed to gain consensus.

> I think, Martin was trying to tell you that VM types have merits and
> are particularly easy to use in function interfaces, are very
> efficient and allow for bounds checking. The way that VLAs are
> referred to in C++ documents (for example the document that originated
> this thread) shows that they are not completely understood in the C++
> community.

I don't know what you mean by that either; can you point to a specific
part of P2128
that you're talking about here? I would hesitate to suggest that lack
of understanding
is what causes the failure of adopting VLAs into C++.