John Levon wrote: >> Than why 'I agree with Angus.' ? > > Because the .C file is a different matter, it does not cause > significant false dependencies since nobody #includes C and link > time is not much of an issue
Actually, my design strategy was each .h file should have a corresponding .C file (so that I can find the blighter ;-), so I would prefer either the one solution (lots of .h files with corresponding .C files) or the other (single .h file with single .C file) but not a mix. Do we need the fine-grained control you propose? lyxlib.h #includes <string> only and is negligible effort to parse. -- Angus