I did this once long ago in C. Nibble through your file char by char and change state at the appropriate events. You need to track states like: directive, code, comment, comment_plus, string_literal, char_literal, directive_string, and directive_angle_string (system lib names). You also need to track if you are at the start_of_line, and would need to look ahead a char on occasion and deal with escapes.

As I recall the compilation sequence of trigraphs, preprocessing, and processing did not have any real impact. I just bailed on trigraphs; I have still never seen a live one. Trigraphs may be non-events for your purpose, I had not considered that on my first go-round.

The Right-Thing is to replace a comment with a single space.

Be well,
rir