* bin/ltlcross.cc: Since is_empty() now works with arbitrary
acceptance conditions, calling remove_fin() is not necessary anymore.
* tests/core/ltlcrossce.test: Adjust.
* NEWS: Mention the change.
* spot/twaalgos/genem.cc, spot/twaalgos/genem.hh: Improve the worst
case by not recurring twice into each disjunct individually. Keep
the previous two implementation available and add a function
generic_emptiness_check_select_version() so we can benchmark the
difference.
* tests/python/genem.py: Test the three versions.
* spot/twaalgos/genem.cc: If a disjunct has no unit-Fin to remove the
code should select any Fin occuring in the disjunct, but it was
selecting any Fin occuring in the acceptance condition (made of
disjuncts) instead. This could potentially double the number of
recursive calls.
* spot/twaalgos/car.cc: Don't reserve colors when parity prefix is used
but the condition has no parity prefix and use colors of prefix in
apply_to_Buchi.
Reported by Florian Renkin.
* spot/twaalgos/cleanacc.cc: Do not rewrite
...Fin(0)&(Fin(1)|(Fin(3)&Inf(4)))... as Fin(0)&f when it appears that
{1,3} cannot receive {0}, because {1} is used elsewhere in the
acceptance.
* tests/python/simplacc.py: Add test case.
* spot/twaalgos/relabel.cc: Remove false transitions if
some of the propositions are equivalent to true or false.
* NEWS: Mention the bug.
* tests/core/ltl2tgba2.test: Test it.
* spot/twaalgos/degen.cc (keep_bottommost_copies): Fix intialisation
of bottommost_occurence.
* tests/python/pdegen.py: Add test case sent by Florian Renkin.
doing it too early breaks the "redirect all-accepting transitions"
optimization
* spot/twaalgos/degen.cc: move the original-states composition
code.
* tests/python/pdegen.py: Add test-case from Florian Renkin.
* bin/autfilt.cc: Add a --partial-degeneralize option.
* NEWS: Mention it.
* spot/twaalgos/degen.cc: Do not restrict partial_degeneralize() to
deterministic automata.
* spot/twaalgos/degen.hh: Adjust documentation.
* tests/core/pdegen.test: New test case.
* tests/Makefile.am: Add it.
* tests/python/pdegen.py: Adjust.
* spot/twaalgos/degen.cc: If an all-accepting transition has been
redirected, we need to force purge_unreachable_state() to be called,
otherwise we may have unreachable states.
* tests/python/pdegen.py: Add test case.
* spot/twaalgos/relabel.cc: Remove false transitions if
some of the propositions are equivalent to true or false.
* NEWS: Mention the bug.
* tests/core/ltl2tgba2.test: Test it.
Reported by Florian Renkin
* spot/twaalgos/degen.cc (partial_degeneralize): Update
original-state when partial_degeneralize is executed more
than once in a loop.
* tests/python/pdegen.py: Add test case.
* spot/twaalgos/degen.cc,
spot/twaalgos/degen.hh (is_partially_degeneralizable): New function.
(partial_degeneralize): New version without todegen argument.
* tests/python/pdegen.py: Add test cases.
* spot/twaalgos/degen.cc, spot/twaalgos/degen.hh: Implement this.
Also throw a runtime error in case were todegen does not match any
subformula.
* tests/python/pdegen.py: Add tests.
On these 4 cases, added to pdegen.py, and supplied by Florian Renkin,
partial_degeneralize() is now at least as good as degeneralize_tba(),
and sometimes better. This is achieved as follows: (1) a
propagate_marks procedure is introduced to propagate marks as far as
possible on the automaton (e.g., common outgoing marks can be push
onto the incoming transitions and vice-versa), (2) the
degeneralization order is compute dynamically, and (3) whenever and
fully-accepting transition is taken in the original automaton, the
destination level is chosen to be the highest existing level.
* spot/twaalgos/degen.cc,
spot/twaalgos/degen.hh (propagate_marks_vector, propagate_marks_here):
New functions.
(partial_degeneralize): Improve, as described above.
* tests/python/pdegen.py: Add test cases.
