* wrap/python/ajax/css/ltl2tgba.css: Fix position of the "Send"
button with WebKit. The folding arrow icon had a vertical border
that overlapped with the next line.
* src/tgbaalgos/minimize.cc (minimize_wdba): Fix the Löding
algorithm to use colors. The previous implementation was an
incorrect approximation.
* src/tgbatest/wdba2.test: New file showing two equivalent
formulas that were minimized in automata with different sizes.
* src/tgbatest/Makefile.am: Add it.
* src/kripketest/defs.in (run2): Remove this function. It was
incorrectly trying to run valgrind even when valgrind is not
installed.
* src/kripketest/kripke.test: Simplify and use run().
* src/tgbaalgos/minimize.cc (minimize_dfa): Fix detection of the
last iteration. An extra iteration case could be missed in case
where a split generates only singletons, and yet predecessor
classes need to be refined.
* src/ltlvisit/length.cc: Fix computation for ltl::multop
operator. "a&b&c" was reported with length 3, ignoring the
"&" operators, because of a typo.
* src/ltlvisit/length.hh: Fix description to correctly
reflect this change intended since 2010-01-22.
* src/ltltest/length.test, src/ltltest/length.cc: New files.
* src/ltltest/Makefile.am: Add them.
There is no point in degeneralizing an automaton if it can be WDBA
minimized. Doing so will only augment the number of states and
slow down the powerset construction used by the WDBA minimization.
* src/tgbatest/babiak.test: New file. It includes 5 formulae
which Tomáš Babiak reported Spot 0.7.1 would take over one hour to
translate if degeneralization and WDBA minimization were both
requested.
* src/tgbatest/Makefile.am (TESTS): Add it.
* src/tgbatest/ltl2tgba.cc: Do WDBA minimization before
degeneralization. The above formulae are now all translated in a
few seconds.
The previous setup failed with clang++ 3.0.
* m4/stl.m4: New file.
* configure.ac: Call AC_HEADER_UNORDERED_MAP,
AC_HEADER_TR1_UNORDERED_MAP, and AC_HEADER_EXT_HASH_MAP.
* src/misc/hash.hh: Include _config.h, and used the
SPOT_HAVE_UNORDERED_MAP, SPOT_HAVE_TR1_UNORDERED_MAP,
or SPOT_HAVE_EXT_HASH_MAP defines to decide which
file to include.
* src/kripketest/Makefile.am (LDADD): Remove a broken dependency.
Reported by Yann Thierry-Mieg.
* src/sanity/style.test: Make sure it does not appear again.
* src/misc/bddalloc.cc (bdd_allocator::initialize): Call
bdd_setmaxincrease(500000), because the default is 50000,
which cause garbage collection to occur too often.
* src/tgbaalgos/neverclaim.cc, src/tgbaalgos/dotty.cc,
src/tgbaalgos/save.cc: Prefer '\n' over std::endl to speedup I/O.
* src/ltltest/genltl.cc (syntax): Use '\n' too, although it won't
make a big difference.
* bench/emptchk/pml2tgba.pl: Stop checking for version start lines
depending on the Spin version. This check was never always
correct. Reported by Étienne Renault.
* bench/emptchk/formulae.ltl: Do not use + and * in the list of
formulas. Use | and & instead. The * operator was removed on
2010-01-30. Reported by Étienne Renault.
* src/tgbaalgos/randomgraph.hh: Document the fact that adding
acceptance conditions to the graph may generate graphs that do not
have any accepting cycle.
Especially with should write !(p0) and not !p0, because p0 is
usually #define'd by the user and he may have forgotten to quote
the value of the macro. This issue was discovered by Kristin
Yvonne Rozier and diagnosed by Gerard Holzmann.
* src/tgbaalgos/neverclaim.cc (process_link): Call
to_spin_string(..., true) to fully parentheses the string.
* src/tgbatest/neverclaimread.test: Add a test.
* src/tgbaalgos/weight.cc (inc_weight_handler)
(dec_weight_handler): Remove these assertions that require the
loop the be completed, and use break to exit ASAP.
* src/neverparse/neverclaimscan.ll: Make the space between `!' and
`(' optional. This fixes the patch from 2011-02-07 that made this
space mandatory...
* src/tgbatest/neverclaimread.test: Augment test case.
* src/tgba/tgbasafracomplement.cc
(tgba_safra_complement::tgba_safra_complement)
(tgba_safra_complement::succ_iter): Correct the declaration and
use of multiple acceptance conditions.
(state_complement::to_string): Output the L set, not U. The previous
code caused different states to share the same names, causing issues
with the text-based output (state with identical names get merged).
* src/tgba/tgbasafracomplement.hh
(tgba_safra_complement::acceptance_cond_vec_): Adjust type to
store BDDs.
* src/tgbatest/complementation.cc: Implement a new "-b" option
to output automata in Spot's syntax.
* src/tgbatest/complementation.test: Add a test-case supplied
by Martin Dieguez Lodeiro.
* THANKS: Add Martin.
* src/tgba/tgbasafracomplement.cc (print_safra_tree): Fix output in
case of hash collision. Use the actual states to get a number, not
their hash value.
(print_safra_automaton): Output a mapping of values to states names.
(safra_tree_automaton::get_sba): New method, used by
print_safra_automaton.
* src/evtgbaalgos/tgba2evtgba.cc (process_link): Fix prototype
to match tgba_reachable_iterator::process_link.
* src/ltlvisit/tunabbrev.hh: Add using super::visit, so that the
other visit() method are in scope when we overload one.
* src/tgba/tgbareduc.hh, src/tgba/tgbareduc.cc (start, end,
process_link): Remove these empty methods. The default
implementations are empty too, and process_link had the
wrong prototype.
* src/tgbaalgos/reductgba_sim.hh, src/tgbaalgos/reductgba_sim.cc
(start, end, process_link): Likewise.
Spot 0.7.1 used to need 190 acceptance conditions to translate the
188 literature formulae. With this patch we are down to 185.
That's not an impressive, but there are only ~20 formulae that
require more than 1 acceptance conditions; hence little room for
improvement.
* src/misc/bddlt.hh (bdd_hash): New function.
* src/misc/accconv.hh, src/misc/accconv.cc: New files.
* src/misc/Makefile.am: Add them.
* src/tgbaalgos/scc.cc (scc_map::build_map): Adjust
to record all combination of acceptance conditions occurring in a SCC.
* src/tgbaalgos/scc.hh (scc_map::scc::useful_acc): Update description.
* src/tgbaalgos/sccfilter.cc (scc_filter): Simplify acceptance
conditions that are always implied by another acceptance
conditions. Previously, we only removed acceptance conditions
that where always present in accepting SCCs.
* src/tgbatest/sccsimpl.test: New file.
* src/tgbatest/Makefile.am (TESTS): Add it.
This avoids a small regression on the size of degeneralized
automata of our usual list of literature formulae.
* src/tgba/tgbatba.hh, src/tgba/tgbatba.cc
(tgba_tba_proxy::union_acceptance_conditions_of_original_state):
New method.
* src/tgba/tgbatba.cc (tgba_tba_proxy_succ_iterator): In accepting
states, ignore only the last expected acceptance condition if its
common to all outgoing transitions AND if it is not used by any
outgoing transitions of the destination.
states).
* src/tgbaalgos/tgbatba.cc: When degeneralizing to SBA, remove the
acceptance conditions that are common to all outgoing transitions
of this state. This helps to make the degeneralization
idempotent.
* src/tgbatest/degenid.test: New test case.
* src/tgbatest/Makefile.am: Add it.
* src/tgbaalgos/save.c (process_state): Escape quotes in the
name of source and destination states. This fixes a side bug
in the upcoming degenid.test test case.
after the simulation-reduction.
Report from Tomáš Babiak <xbabiak@fi.muni.cz>.
* src/tgbaalgos/neverclaim.hh (never_claim_reachable): Take
a tgba as input.
* src/tgbaalgos/neverclaim.cc (never_claim_bfs): Call
state_is_accepting() only if this tgba turns out to be
a tgba_sba_proxy. Otherwise check the acceptance of one
outgoing transition as we do in dotty_bfs since 2011-03-05.
* src/tgbatest/ltl2tgba.cc: Do not redegeneralize before
calling never_claim_reachable() if we know the automaton is
degeneralized already.
* src/tgbatest/ltl2tgba.test: Add a test case.
