[Bug target/81644] ICE in rtl_verify_bb_insn, BBRO pass duplicates BB that ends with flow control insn

["ubizjak at gmail dot com" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81644

Uroš Bizjak <ubizjak at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2017-08-02
          Component|rtl-optimization            |target
   Target Milestone|---                         |8.0
     Ever confirmed|0                           |1

--- Comment #2 from Uroš Bizjak <ubizjak at gmail dot com> ---
It turned out we can't simply emit trap insn during epilogue expansion, since
epilogue is expanded on FALLTHRU edge to EXIT BB. Later passes rightfully choke
on control flow insn in BB with FALTHRU exit edge.

I'm testing the following patch:

--cut here--
Index: config/i386/i386.c
===================================================================
--- config/i386/i386.c  (revision 250818)
+++ config/i386/i386.c  (working copy)
@@ -15199,7 +15199,7 @@
   if (ix86_function_naked (current_function_decl))
     {
       /* The program should not reach this point.  */
-      emit_insn (gen_trap ());
+      emit_insn (gen_ud2 ());
       return;
     }

Index: config/i386/i386.md
===================================================================
--- config/i386/i386.md (revision 250818)
+++ config/i386/i386.md (working copy)
@@ -201,6 +201,7 @@
 ])

 (define_c_enum "unspecv" [
+  UNSPECV_UD2
   UNSPECV_BLOCKAGE
   UNSPECV_STACK_PROBE
   UNSPECV_PROBE_STACK_RANGE
@@ -18606,6 +18607,18 @@
 }
   [(set_attr "length" "2")])

+(define_insn "ud2"
+  [(unspec_volatile [(const_int 0)] UNSPECV_UD2)]
+  ""
+{
+#ifdef HAVE_AS_IX86_UD2
+  return "ud2";
+#else
+  return ASM_SHORT "0x0b0f";
+#endif
+}
+  [(set_attr "length" "2")])
+
 (define_expand "prefetch"
   [(prefetch (match_operand 0 "address_operand")
             (match_operand:SI 1 "const_int_operand")
--cut here--