> Although I think it's a bug for .c files which aren't generated to depend
> on something. If you just change it so that the .o's depend on parse-h
> then everything will work right. I think.

The dependency of the .o files on parse.h ought to be figured out by the make
depend step.  I'll bet the motivation for this is to ensure that parse.h
*exists* during the make depend step, by ensuring that "make depend" runs
the parse-h rule to generate parse.h as part of ensuring that lex.c et al
are "up to date".  If so, there has to be a better way to express that...
(I guess "lint" should also depend on parse-h.)