Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84; site brl-tgr.ARPA
Path: utzoo!watmath!clyde!burl!ulysses!allegra!mit-eddie!genrad!panda!talcott!harvard!seismo!brl-tgr!tgr!cottrell@NBS-VMS
From: cottrell@NBS-VMS
Newsgroups: net.lang.c
Subject: Remake Depend
Message-ID: <9507@brl-tgr.ARPA>
Date: Tue, 26-Mar-85 13:59:31 EST
Article-I.D.: brl-tgr.9507
Posted: Tue Mar 26 13:59:31 1985
Date-Received: Thu, 28-Mar-85 02:02:25 EST
Sender: news@brl-tgr.ARPA
Lines: 30

/*
Kim Walden writes about mkmf...
> Unfortunately, mkmf generates wrong dependencies, in most cases,
> when generated files are involved.

It would probably help to manually make the program first, so that all
needed files exist.
 
> I would hardly call standard use of yacc or lex as 'fancy' by
> any meaning of the word.

The first program a unix novice learns is probably cat or ls. The last
is most likely yacc. Also, most files get compiled from j.x to j.o,
(where j is the name and x is the language suffix) then collected with
other .o's & maybe a few libraries to form one or more executables.
Yacc & Lex go thru an extra step. Even make gets confused about this.
 
> Almost correct is not good enough when hundreds or thousands of
> dependencies are generated.  If you have to make manual corrections
> after each dependency generation, the whole point is missed.

Okay, the alternative is do it by hand. Half a job is better than none.
You can't automate EVERYTHING! Most of the dogwork (of doing it by hand) 
is probably tracing #included files. Default suffix rules & special
cases should be a small part. No program can automatically know which
libraries to include or what ld options to use by looking at the source.
I have never used mkmf, I am commenting on your comments in general.

	jim		cottrell@nbs
*/