ltlsynt: rework synthesis algorithms

ltlsynt now offers two algorithms: one where splitting occurs before
determinization (the historical one) and one where determinization
occurs before splitting.

* bin/ltlsynt.cc: here
* tests/core/ltlsynt.test: test it and refactor test file
* NEWS: document it
* spot/misc/game.hh, spot/misc/game.cc: remove Calude's algorithm

tests/core/ltlsynt.test

@@ -104,7 +104,11 @@ diff out exp
 cat >exp <<EOF
 translating formula done
 split inputs and outputs done
+automaton has 9 states
 determinization done
+dpa has 14 states
+simulation done
+dpa has 14 states
 parity game built
 EOF
 ltlsynt --ins='a' --outs='b' -f 'GFa <-> GFb' --verbose --realizability 2> out
@@ -179,10 +183,11 @@ for i in 0 1 7 8 9; do
   IN=$(eval echo \$IN$i)
   OUT=$(eval echo \$OUT$i)
   EXP=$(eval echo \$EXP$i)
-  test $EXP = $(ltlsynt -f "$F" --ins="$IN" --outs="$OUT" --realizability \
-  --algo=rec)
-  test $EXP = $(ltlsynt -f "$F" --ins="$IN" --outs="$OUT" --realizability \
-  --algo=qp)
+
+  for algo in sd ds; do
+    test $EXP = $(ltlsynt -f "$F" --ins="$IN" --outs="$OUT" --realizability \
+    --algo=$algo)
+  done
 done
 
 for i in 2 3 4 5 6 10; do
@@ -190,11 +195,18 @@ for i in 2 3 4 5 6 10; do
   IN=$(eval echo \$IN$i)
   OUT=$(eval echo \$OUT$i)
   EXP=$(eval echo \$EXP$i)
-  ltlsynt -f "$F" --ins="$IN" --outs="$OUT" > out$i
-  REAL=`head -1 out$i`
-  test $REAL = $EXP
-  tail -n +2 out$i > res$i
+
   ltl2tgba -f "!($F)" > negf_aut$i
-  # check that the L(strategy) is included in L(F)
-  autfilt -q -v --intersect=negf_aut$i res$i
+
+  # test ltlsynt
+  for algo in sd ds; do
+    ltlsynt -f "$F" --ins="$IN" --outs="$OUT" --algo=$algo > out$i || true
+    REAL=`head -1 out$i`
+    test $REAL = $EXP
+    tail -n +2 out$i > res$i
+    # check that the L(strategy) is included in L(F)
+    autfilt -q -v --intersect=negf_aut$i res$i
+    # check that all environment actions are possible
+    autfilt --remove-ap="$OUT" res$i | autfilt --dualize | autfilt --is-empty -q
+  done
 done