Am Dienstag, den 11.08.2020, 09:43 -0700 schrieb Richard Smith:
> On Tue, 11 Aug 2020, 09:16 Uecker, Martin, <
> Martin.Uecker_at_med.uni-goettingen.de> wrote:
>
> > Am Dienstag, den 11.08.2020, 09:04 -0700 schrieb Richard Smith:
> > > On Mon, 10 Aug 2020, 23:37 Uecker, Martin via Liaison, <
> > > liaison_at_[hidden]> wrote:
> >
...
> > > But what about this:
> > >
> > > {
> > > if (a)
> > > label:
> > > }
> > >
> > > Does the if-statement end after the label? Or is that still a parse
> >
> > error?
> > > Neither option seems particularly nice to me.
> >
> > This is still not allowed. Why is this not nice?
> >
>
> It certainly seems like the better option. But it also makes a corner case
> more complex: we go from the rule we've been teaching people for decades
> that "you always need a statement after a label; use ';' if you don't have
> one" to "you sometimes need a statement and sometimes don't". (You can
> certainly argue that it's the if statement that needs a statement after it,
> not the label, but however you describe it the rule seems to end up more
> complex than before.)

I think this misses the point.

In real world we had to explain this rule a lot because
people always wanted to place labels where they wanted
to jump to, but then it turned out that it was often
not allowed by the language for reasons people could
not understand. So we need to teach them how to insert
pointless null statements. With the changed rule this
is simply not necessary anymore.


> Given that the value proposition is fairly weak to start with, especially
> since C++ is on a path away from all constructs involving labels (with the
> heavy discouragement of goto and with the upcoming addition of
> pattern-matching expected to replace switch), the added complexity makes
> this a very borderline change for me.

For me personally the value proposition of making the
change to C++ is zero, as I stopped using C++.


> I think it's still worth asking the C++ committee if they want it. C++
> implementations are presumably going to allow it regardless now it's part
> of C2x, so we may as well standardize it even if we think use of labels is
> on the way out in idiomatic C++.

Keeping the two language closer seems worthwhile, but if labels
are on their way out, it may be a
waste of time...

Best,
Martin



> The following code is still allowed:
> >
> > if (a)
> > label: foo();
> >
> >
> > Best,
> > Martin
> >
> >
> > > Another (old) difference between C is that in C++ you
> > > > can use declarations everywhere where statements are
> > > > allowed, which includes with iteration, control, and
> > > > jump statements.
> > > >
> > > > if (1)
> > > > int i;
> > > >
> > > > This does not make too much sense in the context
> > > > of C, so I do not think we will adopt this.
> > > >
> > > > Best,
> > > > Martin
> > > > _______________________________________________
> > > > Liaison mailing list
> > > > Liaison_at_[hidden]
> > > > Subscription: https://lists.isocpp.org/mailman/listinfo.cgi/liaison
> > > > Searchable archives: http://lists.isocpp.org/liaison/2020/08/index.php
> > > >