[Bug tree-optimization/87896] [9 Regression] ICE in verify_flow_info failed

["rguenth at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87896

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2018-11-06
     Ever confirmed|0                           |1

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
Looks like a latent issue?  We have EDGE_IGNORE set on edges so

          if ((e->flags & ~(EDGE_DFS_BACK
                            | EDGE_CAN_FALLTHRU
                            | EDGE_IRREDUCIBLE_LOOP
                            | EDGE_LOOP_EXIT
                            | EDGE_CROSSING
                            | EDGE_PRESERVE)) == 0)
            n_branch++;

doesn't count them.  The flag is set by
vr-values.c:simplify_switch_using_ranges
but never reset.

/* An edge we should ignore.  It should be entirely local to
   passes.  ie, it is never set on any edge upon the completion
   of any pass.  */
DEF_EDGE_FLAG(IGNORE, 17)

We could clear the flag in vr_values::~vr_values, but not sure if that's
the best thing to do.  VRP and DOM could do that as well.  tree-vrp.c
used to clear EDGE_IGNORE but the issue must be that we somehow _do_
thread (and copy) those edges?!

That is, we seem to look at EDGE_IGNORE only in thread_outgoing_edges but
appearantly not on all edges included in a threading path?  As said edges
marked this way are determined to be never executed.

Jeff?