clockvector: bugfix - bad clock merge

The clock merge function had a really bad bug, where it would copy
garbage into the new clock vector, if the earlier clock vector (cv) was
"wider" than the current clock vector (this).

Now, ClockVector::merge is much simpler and much less buggy.

diff --git a/clockvector.cc b/clockvector.cc
index 062ac4148e6573a4f00f59ccd6b947c5d37259e7..740b5cf2793de217eea47342e5a785537d016362 100644 (file)
--- a/clockvector.cc
+++ b/clockvector.cc
@@ -40,27 +40,15 @@ ClockVector::~ClockVector()
  */
 void ClockVector::merge(const ClockVector *cv)
 {
-       modelclock_t *clk = clock;
-       bool resize = false;
-
        ASSERT(cv != NULL);
 
-       if (cv->num_threads > num_threads) {
-               resize = true;
-               clk = (modelclock_t *)snapshot_malloc(cv->num_threads * sizeof(modelclock_t));
-       }
+       if (cv->num_threads > num_threads)
+               clock = (modelclock_t *)snapshot_realloc(clock, cv->num_threads * sizeof(modelclock_t));
 
        /* Element-wise maximum */
-       for (int i = 0; i < num_threads; i++)
-               clk[i] = std::max(clock[i], cv->clock[i]);
-
-       if (resize) {
-               for (int i = num_threads; i < cv->num_threads; i++)
-                       clk[i] = cv->clock[i];
-               num_threads = cv->num_threads;
-               snapshot_free(clock);
-       }
-       clock = clk;
+       for (int i = 0; i < cv->num_threads; i++)
+               if (cv->clock[i] > clock[i])
+                       clock[i] = cv->clock[i];
 }
 
 /**