From: ian <ian>
Date: Thu, 15 May 2008 23:00:33 +0000 (+0000)
Subject: sort out comments, and elapsed time calculation, for new regime
X-Git-Url: http://www.chiark.greenend.org.uk/ucgi/~ijackson/git?a=commitdiff_plain;h=be6ac63af68570be8095dd1bd4e1ae72e0325e4b;p=trains.git

sort out comments, and elapsed time calculation, for new regime
---

diff --git a/hostside/safety.c b/hostside/safety.c
index e84ffd9..73f1048 100644
--- a/hostside/safety.c
+++ b/hostside/safety.c
@@ -35,18 +35,18 @@ Segment *segments;
  * using the foredetect; when it enters a new segment we advance first
  * the head and then the tail by the same distance.
  *
- * We advance foredetect firstly exactly one segment and then over the
- * current stopping distance.  At each advancement we record the time
- * from the currently-next detection at which this prediction will
- * happen (computed from the current speed).  (We don't take into
- * account any already commanded but not necessarily happened
- * deceleration; however if we are attempting to increase our speed,
- * current speed refers to the new speed.)  If we're already stopping
- * then we advance by the stopping distance only (since we don't need
- * to worry about not being told to stop until the next detection);
- * also we stop advancing when we run out of segments we own (this
- * deals with segment-sized rounding errors which might otherwise
- * make us think we will overrun).
+ * We advance foredetect by the total of the uncertainty (which takes
+ * it into the next segment) and the current stopping distance.  At
+ * each advancement we record the earliest time from now at which this
+ * prediction will happen (computed from the current speed).  (We
+ * don't take into account any already commanded but not necessarily
+ * happened deceleration; however if we are attempting to increase our
+ * speed, current speed refers to the new speed.)  If we're already
+ * stopping then we advance by the stopping distance only (since we
+ * don't need to worry about not being told to stop until the next
+ * detection); also we stop advancing when we run out of segments we
+ * own (this deals with segment-sized rounding errors which might
+ * otherwise make us think we will overrun).
  *
  * At each new track location for our front end we hope to find a
  * segment which no-one currently owns and which is currently set for
@@ -98,7 +98,8 @@ Segment *segments;
  * movfeat changes.
  *
  * If not then we issue an signal stop command (we'd better not be
- * doing that already!) or countermand the speed increase or whatever.  */
+ * doing that already!) or countermand the speed increase or whatever.
+ */
 /*
  * Here is how we use the stuff in the Segment:
  *
@@ -240,7 +241,6 @@ typedef struct {
     accelerating:1,
     stopping:1,
     done_first_new_fdet:1,
-    count_time:1,
     usecurrentposn:1, /* for pred_getmovpos */
     optimistic:1, /* for autopoint */ 
     walk_compute_polarise:1, /* nose_nextseg still needs to worry */
@@ -316,7 +316,7 @@ static void pred_callback_debug(const char *what, TrackLocation *t,
   
   oprintf(DUPO("safety") " predict   %s"
 	  " %c%s dist=%-4d until=%-4ld %c%c%c.%c (was %s%s..%d dist=%-4d)"
-	  "  %c%c%c%c.%c%c%c%c"
+	  "  %c%c%c.%c%c%c%c"
 	  " elapsed=%ld nit=%d,%d\n",
 	  what,
 
@@ -338,7 +338,6 @@ static void pred_callback_debug(const char *what, TrackLocation *t,
 	  u->was_distance,
 	  
 	  "-2"[ u->done_first_new_fdet ],
-	  "-t"[ u->count_time ],
 	  "-c"[ u->usecurrentposn ],
 	  "-o"[ u->optimistic ],
 	  
@@ -462,6 +461,8 @@ static int nose_nextseg(TrackLocation *t, TrackAdvanceContext *c,
     goto movement_ok;
   }
   /* We have a plan but it involves some motion: */
+  if (u->elapsed < 0)
+    return predict_problem(u, t->seg, "arrived but route not set");
   if (t->seg->pred_vacated) {
     if (!t->seg->now_present)
       /* And we're not even there yet!  This is too hard because we
@@ -489,7 +490,7 @@ static int nose_nextseg(TrackLocation *t, TrackAdvanceContext *c,
   if (!t->seg->pred_vacated)
     t->seg->tr_backwards= t->backwards;
   t->seg->pred_present= 1;
-  t->seg->until= u->elapsed;
+  t->seg->until= u->elapsed < 0 ? 0 : u->elapsed;
   return 0; /* yay! */
 }
 
@@ -508,7 +509,7 @@ static int tail_nextseg(TrackLocation *t, TrackAdvanceContext *c,
   if (before->seg->pred_vacated) return 0; /* only vacate once */
   before->seg->pred_present= 0;
   before->seg->pred_vacated= 1;
-  before->seg->until= u->elapsed;
+  before->seg->until= u->elapsed < 0 ? 0 : u->elapsed;
 
   return 0;
 }
@@ -530,9 +531,7 @@ static int fdet_nextseg(TrackLocation *t, TrackAdvanceContext *c,
   if (!before) return 0;
 
   advanced= calc_advanced(c);
-
-  if (u->count_time)
-    advance_elapsed(u,advanced);
+  advance_elapsed(u,advanced);
 
   u->nosec.distance= advanced;
   r= trackloc_advance(&u->nose,&u->nosec);
@@ -605,9 +604,6 @@ static int fdet_nextseg(TrackLocation *t, TrackAdvanceContext *c,
   /* Final adjustments, prepare for next iteration */
 
   u->was_distance= c->distance;
-  u->count_time= 1; /* we start counting time only after we've done the
-		     * whole foredetect segment as we may already have
-		     * nearly finished it */
   return 0;
 }
 
@@ -717,7 +713,6 @@ ErrorCode predict(Train *tra, int accelerate, struct timeval tnow,
   u.nosec.nextseg= nose_nextseg;
   u.nosec.trackend= pred_trackend;
   u.nosec.distance= nose_length(tra);
-  u.count_time= 0;
 
   u.nose= u.fdet;
 
@@ -736,6 +731,7 @@ ErrorCode predict(Train *tra, int accelerate, struct timeval tnow,
     u.fdetc.distance= u.stopping_distance;
   } else {
     u.fdetc.distance= tra->uncertainty + u.stopping_distance;
+    u.elapsed= -tra->uncertainty / u.maxspeed;
   }
   u.fdetc.distance += MARGIN_NOSE;
   u.was_distance= u.fdetc.distance;