[Bug fortran/125747] ICE in gfc_add_modify_loc, at fortran/trans.cc:229 since r16-5177

[kargl at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125747

--- Comment #8 from Steve Kargl <kargl at gcc dot gnu.org> ---
(In reply to Harald Anlauf from comment #7)
> (In reply to Steve Kargl from comment #6)
> > (In reply to Harald Anlauf from comment #4)
> > > 
> > > I think we should not allow this extension for modern DO constructs.
> > >
> > 
> > I think it should be disallowed in FORALL as well.  The error
> > message in resolve.cc(gfc_resolve_iterator) is guarded with 
> > gfc_notify_std (GFC_STD_F95_DEL...).   F95 never allowed REAL 
> > iterators and FORALL was introduced in this standard.
> > 
> > F95, p.114
> > 
> > R750  forall-triplet-spec  is index-name = subscript : subscript [ : stride]
> > R617  subscript            is scalar-int-expr
> > R603  stride               is scalar-int-expr
> > 
> > If it breaks code in the wild, so what?  The code was never correct.
> 
> +1
> 
> When I specify -std=f2023,
> 
>   do j = 1, 0.3
>   end do

gfortran had to accept this one due to f66/f77/g77 compatibility.
I had a hand in getting that to work.  With -std=f2023, one gets
the error, without -std=f2023 a warning is issues, and with 
-std=legacy it's silently accepted.

> is already rejected, and
> 
>   forall (j = 1:0.3)
>   end forall
> 
> anyway.  So it is actually a regression...