Several bugs have been found where the MallocSiteTable is overflowed. The index to the table might be 128 or some non-prime number. This should be improved.
Comments
Found the other bug that caused this, we'll not change the hash. Closing as WNF
09-03-2015
// Currently, (number of buckets / number of entires) ratio is
// about 1 / 6
enum {
table_base_size = 128, // The base size is calculated from statistics to give
// table ratio around 1:6
table_size = (table_base_size * NMT_TrackingStackDepth - 1)
};
NMT_TrackingStackDepth = 4, so the table size is 511, which is divisible by 7.
ATM we only have stress tests that overflow the table though.