Wireless sensor network applications run on resource scarce sensor nodes. A sensor node usually has extremely limited resources in order to run long time with limited power supply to sense its environment. For example, a XSM[6] sensor node has only 4KB flash memory, a 4MHz processor and a low power radio device.
Sensor network applications are different from computer applications. Most sensor network applications are event driven to run longer with limited power supply. They are in passive mode or sleeping in most of time and can only be activated by some events.
It is very hard to debug sensor network applications because (1) the event driven character of sensor network application makes it hard to predict program behavior; (2) sensor network usually run batched jobs which means people cannot run or debug wireless sensor applications interactively; (3) sensor network application bugs can only be detected after the software is deployed because of the highly distributed and faulty nature of sensor nodes;(4) there is extremely limited of visibility to the running program.
This thesis proposals a source-code based dynamic analysis approach to debug wireless sensor network applications. The NesC compiler is modified to automatically insert data collecting code at each program point where program control flow changes. When wireless sensor network applications run on sensor nodes, the reached program point information will be automatically recorded. When program finished, the collected data can show exactly how program runs. The benefit of this approach is that the data collected are from program product run, it shows the program behavior in real running environment. So the bugs found here are more valuable.