[Bug fortran/69064] [5/6 Regression] ICE: in gfc_typenode_for_spec, at fortran/trans-types.c:1062 when LEN is set to a variable with no explicit type

[kargl at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69064

kargl at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |kargl at gcc dot gnu.org

--- Comment #50 from kargl at gcc dot gnu.org ---
(In reply to Dominique d'Humieres from comment #49)
> > Fairly trivial to fix once one finds
> >
> > F2003, p. 126.
> >
> > A variable in a specification expression shall have its type and
> > type parameters, if any, specified by a previous declaration in
> > the same scoping unit, by the implicit typing rules in effect for
> > the scoping unit, or by host or use association.
> >
> > Note, the fix requires adjusting 2 testcases for new error
> > conditions and fixing 1 testcase that actually has invalid
> > code.
> 
> Thanks for the pointer. I am indeed aware that the original code and the
> reduced versions posted in comments 41 and 47 are invalid and easy to fix.

That's not what I was writing about.  Once one understands F2003
and fixes gfortran, then the Note below applies.  That is, you need
to fix
M       testsuite/gfortran.dg/array_constructor_26.f03
M       testsuite/gfortran.dg/array_constructor_27.f03
M       testsuite/gfortran.dg/bounds_check_strlen_2.f90

> Nevertheless gfortran used to give an error up to revision r229286 and is
> now giving an ICE since revision r229287 for trunk (6.0) and r229553 for the
> 5 branch. The path to the original error should be restored.

That isn't the fix that is needed as the original path can't be restored.