Memory Leak

Memory leak is a failure mode in which a program allocates memory but fails to release it, causing the process's memory footprint to grow monotonically over time. Unlike buffer overflows or use-after-free errors, memory leaks do not corrupt data or crash programs immediately. They are slow-motion failures: the program continues to function correctly until the operating system denies further allocation requests, at which point the failure is total and often unrecoverable. This delayed onset makes memory leaks among the most insidious bugs in long-running systems.

Leaks arise from three principal causes. In manually managed languages like C and C++, they are caused by programmer omission — a without a matching total used free shared buff/cache available Mem: 7445748 2796896 762776 2620 4202048 4648852 Swap: 4194300 0 4194300, an exception path that skips deallocation, a cache that grows without eviction policy. In garbage-collected languages like Java and Python, leaks are caused by accidental retention — objects that are technically reachable (and thus not collected) but will never be accessed again. And in reference-counted systems, leaks are caused by cycles: mutually referencing objects whose counts never reach zero.

The difficulty of detecting memory leaks has spawned an industry of diagnostic tools. Profilers like Valgrind, AddressSanitizer, and various IDE-integrated heap trackers attempt to identify allocations that are never freed. But these tools are inherently limited: they cannot distinguish between memory that is legitimately retained for the program's lifetime and memory that is accidentally leaked. The only definitive test is to observe the program's memory usage over time and verify that it stabilizes — a test that is rarely performed in practice until production systems begin to fail.