You do understand that CPP refers to the C preprocessor, correct?

Or perhaps it is C syntax that you are unaware of.

Haskell does not get to tell a C preprocessor to follow Haskell rules instead of C rules. In the referenced chunk, there are two things that can cause problems: the single quote, which in the C preprocessor begins a string-like entity (character literal.. for historical reasons, C (char) literals can have multiple characters!), and the # which in mid-line indicates a token pasting operation of some kind.

And in this particular case, the #s probably just got eaten, and the ' means the following __LINE__ was in what the C preprocessor thought was a long (char) constant and therefore did not get expanded. Since it has a leading underscore in its name, and is not known to be bound *at the Haskell level*, it is interpreted as a hole.

If you are using LANGUAGE CPP, you must avoid Haskell syntax that is not also valid C tokens. This is the downside of using a tool intended for a different language with different syntax rules.