fixed bug: ticks_since_last_event wasn't getting reset on button hold release

The visible symptom was: Ramp up for 1s or longer, release, wait more than 1s,
then hold again.  It should ramp up, but it would ramp down instead.  The
clause for resetting ramp_direction wasn't happening, because the EV_tick
counter started at a value higher than 1s where it would normally trigger.

The underlying cause was a bit complicated.  Recent changes in PCINT_inner()
were causing ticks_since_last_event to get set to 0 (at push_event()) and then
back to its previous value (at emit_current_event()).  The EV_tick counter would
then start at whatever the button release event used.

The fix involved removing the part of emit_current_event() where it would
set ticks_since_last_event to "arg".  That line was a very old bug which simply
hadn't caused any visible issues until recently.  Instead, it needs to set
ticks_since_last_event more carefully, at other locations.  Specifically, it
resets to 0 now in empty_event_sequence() and one more location in the deferred
WDT handler (when HOLD_TIMEOUT triggers).

Additionally, push_event() was only ever used from PCINT_inner()...  so I moved
the tick reset logic to PCINT_inner() instead.  This allows us to decrease size
by about 10 bytes, since PCINT_inner() no longer needs to copy the counter before
it gets reset.  However, it also means push_event() should never be called from
any other function.

1 files changed, 2 insertions, 2 deletions

diff --git a/spaghetti-monster/fsm-wdt.c b/spaghetti-monster/fsm-wdt.c
index 94266c1..9e8d9af 100644
--- a/spaghetti-monster/fsm-wdt.c
+++ b/spaghetti-monster/fsm-wdt.c
@@ -150,6 +150,7 @@ void WDT_inner() {
         // (first frame of a "hold" event)
         else {
             if (ticks_since_last >= HOLD_TIMEOUT) {
+                ticks_since_last_event = 0;
                 current_event |= B_HOLD;
                 emit_current_event(0);
             }
@@ -160,9 +161,8 @@ void WDT_inner() {
     else if (current_event) {
         // "hold" event just ended
         // no timeout required when releasing a long-press
-        // TODO? move this logic to PCINT() and simplify things here?
         if (current_event & B_HOLD) {
-            //emit_current_event(0);  // should have been emitted by PCINT_inner()
+            //emit_current_event(ticks_since_last);  // should have been emitted by PCINT_inner()
             empty_event_sequence();
         }
         // end and clear event after release timeout