[Bug tree-optimization/84740] [8 Regression] ICE in build_constructors, at tree-switch-conversion.c:965

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84740

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jakub at gcc dot gnu.org
           Assignee|marxin at gcc dot gnu.org          |jakub at gcc dot gnu.org

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Untested fix:
--- gcc/tree-switch-conversion.c.jj     2018-01-09 08:58:15.000000000 +0100
+++ gcc/tree-switch-conversion.c        2018-03-07 12:03:42.228246408 +0100
@@ -1563,7 +1563,8 @@ process_switch (gswitch *swtch)
   gather_default_values (info.default_case_nonstandard
                         ? gimple_switch_label (swtch, 1)
                         : gimple_switch_default_label (swtch), &info);
-  build_constructors (swtch, &info);
+  if (info.phi_count)
+    build_constructors (swtch, &info);

   build_arrays (swtch, &info); /* Build the static arrays and assignments.  
*/
   gen_inbound_check (swtch, &info);    /* Build the bounds check.  */

I think it is better to do this rather than if (info.phi_count) return "no
non-virtual phis"; before create_temp_arrays, because this way we optimize the
switch into if, while otherwise we don't.  Most of the functions will not
really do much, e.g. allocate 0 elts arrays, etc., but only build_constructors
relies on info->phi_count > 0.