Discovering Heap Anomalies in the Wild

Опубликовано 6 сентября 2016, 16:44

Despite the best efforts of programmers, programs still ship with bugs.  Many of these bugs manifest as anomalies the heap.  This talk discusses two low-overhead synergistic techniques for discovering heap anomalies by exploiting the underlying runtime system.  The first, dynamic object sampling, is a technique for selectively tagging objects with characteristics.  Summarization graphs provide a compact representation of collecting heap characteristics.  We show that these techniques can stand alone or work together to provide ways to discover heap-based bugs with low enough overhead to consider using them in production systems.