merge

When adapting the code to handle transition-based acceptance, a bug was
introduced so that only accepting edges were copied to the result
automaton.

The self-loop labeled as false edge-case is here as prevention, I am not
sure it happens in practice.

* spot/twaalgos/translate_aa.cc: Here.

spot/twaalgos/translate_aa.cc

@@ -124,19 +124,16 @@ namespace spot
                 bool edge_is_acc = ((trans_based && e.acc)
                                     || (!trans_based && sere_aut->state_is_accepting(e.dst)));
 
-                if (edge_is_acc)
+                unsigned new_edge;
-                  {
                 // point to accepting sink instead of original dst if asked
-                    if (use_accepting_sink)
+                if(use_accepting_sink && edge_is_acc)
-                      aut_->new_edge(new_st, accepting_sink_, e.cond);
+                  new_edge = aut_->new_edge(new_st, accepting_sink_, e.cond);
                 else
-                      {
+                  new_edge = aut_->new_edge(new_st, register_state(e.dst), e.cond);
-                        unsigned new_e = aut_->new_edge(new_st, register_state(e.dst), e.cond);
+
                 // remember if old edges were accepting
-                        if (acc_edges != nullptr)
+                if (acc_edges != nullptr && edge_is_acc)
-                          acc_edges->push_back(new_e);
+                  acc_edges->push_back(new_edge);
-                      }
-                  }
               }
           }
 
@@ -340,6 +337,9 @@ namespace spot
             for (unsigned i : acc_edges)
               {
                 auto& e1 = aut_->edge_storage(i);
+                // self loop used to mark accepting state, skip it!
+                if (e1.cond == bddfalse)
+                  continue;
                 for (const auto& e2 : aut_->out(rhs_init))
                   aut_->new_edge(e1.src, e2.dst, e1.cond & e2.cond);
               }