intricate memory corruption bug prevented me to check the patch
in. It took me two days to realize why find_state() must do a
double loop over the candidates to check for equality before
checking for inclusion(s).
* iface/gspn/ltlgspn.cc: New options, -e45 and -n.
* iface/gspn/ssp.cc, iface/gspn/ssp.hh: Handle these.
* src/tgbaalgos/gtec/gtec.cc (TRACE): Add some debugging traces.
(couvreur99_check_shy::dump_queue): New function.
* src/tgbaalgos/gtec/gtec.hh (couvreur99_check_shy::dump_queue):
New function.
* src/ltlvisit/basicreduce.cc (basic_reduce_visitor::visit):
Simplify the formula again after FX->XF and GX->XG permutations.
This is so that formulae like GFXXa become GFa and not just GFXa.
* src/ltlvisit/contain.cc (reduce_tau03_visitor): Fix a typo
in the rules for i|j or i&j, resulting in missing simplifications.
* src/ltlvisit/reduce.hh, src/ltlvisit/reduce.hh:
Add Reduce_Containment_Checks and Reduce_Containment_Checks_Stronger
flags, and call reduce_tau03.
* src/ltlvisit/contain.hh (reduce_tau03): Make "stronger" the
default.
* src/ltlvisit/contain.cc: Style.
* src/ltltest/reduc.cc: Simplify using the reduce() interface
instead of reduce_tau03.
* src/tgbatest/ltl2tgba.cc: Likewise. Add -fr5, -fr6, and -fr7
options.
* src/tgbatest/spotlbtt.test: Remove cases using "-c", since its
current implementation is not always correct (and apparently
reduces less than -fr7).
New function, performing LTL reduction a la tauriainen.03.a83.
* src/ltltest/equals.cc, src/ltltest/reduc.cc: Add support for
the new reduction.
* src/ltltest/reduc.test: Cut the test in half, and additionally
test the new reduction.
* src/ltltest/reduccmp.test: Run on the new reduction.
* src/ltltest/Makefile.am: Adjust.
* src/tgbatest/ltl2tgba.cc: Add new options to apply the reduction.
* src/tgbatest/spotlbtt.test: Use them.
New function.
* src/tgba/tgbaproduct.cc, src/tgba/tgbaproduct.hh: Use it to
distinguish acceptance conditions that are identical in both
operands.
* src/tgbatest/explpro4.test: New file.
* src/tgbatest/explpro2.test, src/tgbatest/Makefile.am: Adjust.
* src/ltlvisit/contain.cc, src/ltlvisit/contain.hh
(spot::ltl::language_containment_checker): ... in these new files.
* src/ltlvisit/Makefile.am: Adjust.
Rename as ...
(free_anonymous_list_of): ... this, and correct their update on
release. Also correct yesterday's the correction (ahem!).
(dump): Improve verbosity.
* src/misc/freelist.cc (free_list::remove, free_list::insert): Fix
longstanding thinkos.
(free_list::free_count): New function.
(ltl_to_tgba_fm): Add a new option "containment_checks" to enable
some language containment checks (via emptiness checks) during the
translation. This first attempt currently only use containment
checks to merge states bisimulating each other.
* src/tgbatest/ltl2tgba.cc: Bind this to option "-c".
* src/tgbatest/spotlbtt.test: Check it.
call release_n(), not remove() to repopulated the freelist of
anonymous BDD variables. New code I'm working on triggered an
assertion inside remove(), but I'm surprised this bug hadn't
manifested before !
(couvreur99_check_shy_ssp): Add a onepass_ attribute to
disable the "shyness", and do not increment pos before calling
find_state since gspn's implementation uses it.
* iface/gspn/ssp.cc: Enable "onepass_" for all "shy" variants,
and also fix find_state for the case where onepass_ is
disabled (but I do not yet know why the latter fix isn't enough).
instead of one : one for the atomic propositions, and one
for the acceptance conditions. This way it's easy for
the tools in iface/gspn/ to require some atomic proposition
to be declared and allow any acceptance conditions (there is nothing
to adjust in this files because of the default value of the argument).
* src/tgbaparse/tgbaparse.yy: Adjust.
* src/tgbatest/ltl2tgba.cc, src/tgbatest/readsave.cc,
src/tgbatest/reductgba.cc, src/tgbatest/tgbaread.cc: Adjust calls.
Reorganize this function so that syntactically there is only one
loop over the successors, and not two. Call reintroduce the call
to couvreur99_check_shy::state_index(), needed by SSP, and
suppress that to index_and_insert introduced on 2004-12-29. Also
split the "group" option in two: "group" and "group2". "group2"
is the equivalent of the older "group", while the new "group" is
weaker and faster.
(couvreur99_check_shy::state_index): Change prototype as needed by
the algorithm.
* src/tgbaalgos/gtec/gtec.hh: Adjust.
* src/tgbaalgos/gtec/nsheap.hh, src/tgbaalgos/gtec/nsheap.cc
(index_and_insert): Remove.
* iface/gspn/ssp.cc (couvreur99_check_shy_ssp::state_index): Adjust
to new prototype.
* bench/emptchk/README, bench/emptchk/algorithms: Adjust references
to group/group2.
(tau03_opt_search): Add option "ordering" (off by default).
If enabled, initialize an explicit ordering for acceptance
conditions into the new member "cond" (a vector of bdds).
(project_acc): New helper function for projecting a set of
acceptance conditions to a subset that maximizes the number
of initial consecutive conditions covered by the set in the
condition ordering.
(dfs_blue): Implement the ordering heuristic.
(dfs_red): Use a sentinel in condition_stack to avoid explicit
checks for the stack's emptiness.
Consider also the conditions in acc when checking for the
completion of an accepting cycle.
Fix the implementation of condition heuristic.
Implement the ordering heuristic.
Simplify the removal of elements from condition_stack (due to
the way in which elements are pushed on the stack, there can
be at most one element with a given depth in the stack at any
one time).
virtual destructors.
* src/tgba/tgbabddfactory.hh (tgba_bdd_factory): Likewise.
* src/misc/hash.hh: Use the std namespace only with GCC 3.0,
not with all compiler versions with minor version 0.
* src/tgba/tgbatba.cc (tgba_tba_proxy_succ_iterator): Fix
friend declaration of ::spot::tgba_tba_proxy.
