[Bug fortran/71883] [5/6/7 Regression] ICE in identical_array_ref, at fortran/dependency.c:104

[kargl at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71883

--- Comment #8 from kargl at gcc dot gnu.org ---
(In reply to paul.richard.tho...@gmail.com from comment #6)
> Hi Steve,
> 
> Thanks, you beat me to it!
> 
> Cheers
> 
> Paul
> 
> PS Since I caused this regression, perhaps I should take it on :-)
> 
;
> > +      {
> > +       int e, w;
> > +       gfc_get_errors (&w, &e);
> > +       if (e < 1)
> > +         gfc_internal_error ("gfc_is_constant_expr(): Unknown "
> > +                               "expression type");
> > +       return 0;
> > +      }
> >      }

I'm tempted to suggest that gfc_internal_error() should be 
modified to check for a nonzero error count.  If gfortran
has already issued one or more errors, then gfc_internal_error
should simply exit; otherwise, issue the error.  Something like

Index: error.c
===================================================================
--- error.c     (revision 238385)
+++ error.c     (working copy)
@@ -1303,10 +1303,15 @@ gfc_error (const char *gmsgid, ...)
 void
 gfc_internal_error (const char *gmsgid, ...)
 {
+  int e, w;
   va_list argp;
   diagnostic_info diagnostic;
   rich_location rich_loc (line_table, UNKNOWN_LOCATION);

+  gfc_get_errors (&w, &e);
+  if (e > 0)
+    exit(EXIT_FAILURE);
+
   va_start (argp, gmsgid);
   diagnostic_set_info (&diagnostic, gmsgid, &argp, &rich_loc, DK_ICE);
   report_diagnostic (&diagnostic);