I am writing a tool that processes some files. This tool will have a command-line interface but can also be used as a class library from third-party code. To deal with an error, I just throw an exception. Third-party code can handle the exception, and the command-line interface might just print it and abort. However, apart from fatal errors, it is also possible that a situation arises that is not fatal and the process can continue, for which I would like to ‘throw a warning’ and continue.
How do I deal with warnings, in such way that both third-party code and the command-line interface can do something with it?
You throw an exception for this as well (assuming the error condition is indeed exceptional – that is expected to be rare).
Exceptions are not always lethal – you need to throw a specific type of exception that the third party code and command line code can catch and act on.
One rule of exception handling is to handle exceptions you know how to handle – if you have a specific exception type for the error and you document it, client code should know how to deal with it (or not).
If what you are trying to do is provide information to using code – warnings that are handled within your library already but that the users may want to know – you can output traces using the tracing subsystem. All users will need to do is configure a listener and will be able to get the information during runtime.