Hi, after some debugging of a higher-level tool I found out that when I use cpphs as a library and the `runCpphs` function that is to produce the preprocessed output, when it comes across the #error directive it will terminate my program. This is because handling #error is implemented with Haskell's `error`. I find that slightly unfortunate since it means I cannot distinguish betwen an #error written in the input file and a programming error inside cpphs. @Malcolm, would you mind a change towards throwing an exception that is different from error so that it can be easily caught, or even better, a change from runCpphs :: ... -> IO String to runCpphs :: ... -> IO (Either String String) or similar? If an exception based interface is kept, it would be nice to add some haddock to `runCpphs`; not knowing about the existence of #error, it is easy to assume that the IO is only used for accessing the FilePath passed in.