Aw: Re: Re: Re: [PATCH] PR fortran/100950 - ICE in output_constructor_regular_field, at varasm.c:5514

Harald Anlauf anlauf@gmx.de
Fri Aug 20 12:17:48 GMT 2021


> Gesendet: Freitag, 20. August 2021 um 14:01 Uhr
> Von: "Tobias Burnus" <tobias@codesourcery.com>
> On 20.08.21 12:53, Harald Anlauf wrote:
> 
> > I played with variations of the testcase by specifying illegal
> > substring bounds, but all these cases were caught in a different
> > spot with similar error messages.
> 
> I can confirm this. – I think in order to reduce the clutter, the
> diagnostic probably should be removed.

I am unable to prove that we will never that check.  So how about:

diff --git a/gcc/fortran/simplify.c b/gcc/fortran/simplify.c
index eaabbffca4d..04722a8640c 100644
--- a/gcc/fortran/simplify.c
+++ b/gcc/fortran/simplify.c
@@ -4552,27 +4552,13 @@ substring_has_constant_len (gfc_expr *e)
 
   if (istart <= iend)
     {
-      char buffer[21];
-      if (istart < 1)
-       {
-         sprintf (buffer, HOST_WIDE_INT_PRINT_DEC, istart);
-         gfc_error ("Substring start index (%s) at %L below 1",
-                    buffer, &ref->u.ss.start->where);
-         return false;
-       }
-
       /* For deferred strings use end index as proxy for length.  */
       if (e->ts.deferred)
        length = iend;
       else
        length = gfc_mpz_get_hwi (ref->u.ss.length->length->value.integer);
-      if (iend > length)
-       {
-         sprintf (buffer, HOST_WIDE_INT_PRINT_DEC, iend);
-         gfc_error ("Substring end index (%s) at %L exceeds string length",
-                    buffer, &ref->u.ss.end->where);
-         return false;
-       }
+
+      gcc_assert (istart >= 1 && iend <= length);
       length = iend - istart + 1;
     }
   else

Or skip the gcc_assert and cross fingers... (we then would not even need to
verify ref->u.ss.length in that depth).

Harald



More information about the Gcc-patches mailing list