sm_coretest_forced_finish_1

Core test for the forced finish of an asynchronous client behavior. It is not a reference state machine; it exists to drive one core code path on demand, like sm_coretest_transition_speed_1 next to it.

What it does

StA  ──EvCountdownEnd──▶  StRest  ──EvTimer<CbTimerCountdownLoop>──▶  StA
 │                          │
 └── EvCbSuccess/EvCbFinished<CbSleepFor, OrSleep> ──▶ StTrap (terminal) ◀──┘

• StA runs CbSleepFor(30s) in OrSleep and CbTimerCountdownOnce(1s) in OrTimer, plus an SrEventCountdown(1) reactor on the timer event. The timer always wins, so on every StA exit the state machine thread calls requestForceFinish() on the sleep behavior while its onEntry thread polls isShutdownRequested(). The behavior then posts EvCbSuccess and EvCbFinished with CURRENT_STATE lifetime while the machine is STATE_EXITING, so both hit the discard path in ISmaccStateMachine::postEvent.
• StRest waits 1 s on a CbTimerCountdownLoop. It is deliberately a different timer behavior than StA’s: StA’s own EvTimer<CbTimerCountdownOnce> is still queued behind the reactor’s EvCountdownEnd when StRest is entered and would otherwise fire StRest’s reaction immediately.
• StTrap is terminal and logs FATAL "TRAP ENTERED". Both StA and StRest transition to it on any sleep event. The sleep can never finish within StA’s 1 s lifetime, so a sleep event processed anywhere is one that crossed a state exit instead of being discarded.

One cycle takes 2 s. Expected per cycle: one requestForceFinish, two CURRENT STATE SCOPED EVENT DISCARDED warnings, zero TRAP ENTERED.

Running

ros2 launch sm_coretest_forced_finish_1 sm_coretest_forced_finish_1.launch.py
# output is teed to /tmp/sm_coretest_forced_finish_1_latest.log

Quick check of a log:

L=/tmp/sm_coretest_forced_finish_1_latest.log
echo "cycles: $(grep -ac '\[StRest\] Entered' $L)  discards: $(grep -ac 'EVENT DISCARDED' $L)  trap: $(grep -ac 'TRAP ENTERED' $L)"

Under sanitizers

Build the core and this package into a separate build/install base with -fsanitize=thread,undefined or -fsanitize=address,undefined -fsanitize-recover=address, then run the node binary directly (not through ros2 run with a stdbuf prefix, which preloads a library ahead of the sanitizer runtime). ThreadSanitizer needs setarch x86_64 -R on kernels with 32-bit mmap randomization. For LeakSanitizer to report, send SIGINT while the machine is in StRest, when no behavior thread is alive. The full procedure and the recorded results are in docs/smacc2_core_defect_fix_study.md at the repository root.