Hi David, > 1. Is it possible to store and execute previously > generated offline test? Something like a "replay" function? Tests are stored as test logs. You can generate an offline test by $ fmbt -l offline.log -o adapter=dummy test.conf -o adapter=dummy disables whatever adapter is defined in test.conf. The most robust way to execute the same test steps again is fmbt's interactive mode (fmbt -i). You can execute arbitrary actions by typing them to interactive mode prompt: $ fmbt -l replay.log -i test.conf fMBT> i:step 1 executing: i:step 1 adapter: i:step 1 model: ok ... On the other hand, you can use fmbt-log to extract executed actions from logs: $ fmbt-log -f '$as' offline.log ...thus you can replay test steps from offline.log by: $ fmbt-log -f '$as' offline.log | fmbt -l replay.log -i test.conf > 2. Maybe this one is a bit dumb but i still have the doubt. > Most of the examples I found use the aal adapter but I would like > to have the adapter and the model totally independent ( I might > use the same adapter for a different model). You can write a separate AAL file that implements only adapter blocks for test steps. Example: pure_model.aal: ---8<--- aal "pure_model" { language "python" {} variables { x } initial_state { x = 1 } input "step 1" { guard { return x == 1 } body { x = 2 } } input "step 2" { guard { return x == 2 } body { x = 1 } } } --->8--- pure_adapter.aal: ---8<--- aal "pure_adapter" { language "python" {} variables {} input "step 2" { adapter { print "*** STEP 2" } } input "step 1" { adapter { print "*** STEP 1" } } input "step 3" { adapter { print "*** STEP 3" } } } --->8--- test.conf that uses pure_model for test generation, while tests are implemented in pure_adapter: ---8<--- model = aal_remote(remote_pyaal -l model.log pure_model.aal) adapter = aal_remote(remote_pyaal -l adapter.log pure_adapter.aal) heuristic = lookahead(4) coverage = perm(1) pass = steps(4) --->8--- Note that pure_adapter.aal cannot use the variables defined in pure_model.aal. On the other hand, you can implement a remote adapter with any programming language pretty easily. A reference implementation in Python (adapter.py): ---8<--- import sys LOGFILE = "/tmp/reference-adapter.txt" def log(msg): file(LOGFILE, "a").write("%s\n" % (msg,)) log("started") # Read test step count step_count = int(sys.stdin.readline().strip()) # Read names of test steps steps = [] for i in xrange(step_count): steps.append(sys.stdin.readline().rstrip()) log("%s steps read" % (step_count,)) # Test execution loop while True: # Read test step to be executed try: step_index = int(sys.stdin.readline().rstrip()) except ValueError: break step_name = steps[step_index] # Execute test step # if step_name == "i:foo": # ... # elif ... log("executing %s" % (step_name,)) # Report the result sys.stderr.write("%s\n" % (step_index,)) # report successful execution # sys.stderr.write("0\n") would report failure # sys.stderr.write(some_other_index) would report some other step # Adapter exit log("stopped") --->8--- This is how you can configure a test run that uses the reference adapter: ---8<--- model = aal_remote(remote_pyaal -l model.log pure_model.aal) adapter = remote_noencode(python adapter.py) heuristic = lookahead(4) coverage = perm(2) pass = steps(20) --->8--- remote_noencode communicates with spawned adapter process through stdin/out/error, and according to the remote adapter protocol specified in the Remote adapter section of https://github.com/01org/fMBT/blob/master/doc/adapters.txt (noencode means that names of actions will not be URL encoded as usual). I hope this helps, Antti