Michael is a software engineer for HP IT Management Software and can be contacted at [email protected].
Experienced C/C++ programmers know about the need to properly manage dynamically allocated memory to avoid memory leaks. Unfortunately, many of us still find ourselves in a tough position when there is a memory leak in the software. How is it detected in the first place? Simply looking at the task manager (or other appropriate tool that shows memory-use statistics) indicates that the memory used by the process is constantly growing. That means that the memory use is expected to remain constant. But the program must allocate memory, at least at the beginning of its life. Therefore, a more precise description would be that the process is running for a while, the input rate is constant, and the memory continues to climb. The interpretation of the input rate depends on the purpose of the program; for a web server, for instance, it can be network traffic throughput, or the number of requests per second.
How do you attack this problem? Assume that your first attempts at just looking at the code failed, and you need to get help from an automatic memory-leak detection tool. While your favorite tool may use unique techniques to trace memory allocation/deallocation and different algorithms to organize that information at runtime, it most probably works like this:
- From the moment the program starts, it traces each memory allocation (probably with additional information, such as a call stack).
- It registers all released memory.
- Before the program terminates, it prints out information about each unreleased memory block.