User Stubs

The advantage of normal stubs described above is flexibility. They also don't require any changes in target code. However, their execution is slow, because the target is stopped for each stub hit.
User stubs, on the other hand, are implemented as modification of the stubbed function on the target. Modifications can be of two types:

Call to other function on target is put at the start of the stubbed function. The other function is called replacement or stub function, and is usually written specifically for testing purposes. It is not used during normal execution of the program.
Return instruction is put at the stubbed function start. This means that return value and output parameters are not set, so this approach is generally useful only for functions returning void and without output parameters or other side effects, like modifications of global variables. These type of stubs are called skip stubs, because they skip function call.

Note that both modifications mentioned above are performed in object code on the target by testIDEA, so you don't have to rebuild and download the code. After test the original object code is restored.

Note: It is important that prototypes of the stubbed and replacement functions match, otherwise results can be unpredictable.