Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!rutgers!tut.cis.ohio-state.edu!att!ulysses!andante!alice!andrew
From: andrew@alice.UUCP (Andrew Hume)
Newsgroups: comp.unix.wizards
Subject: Re: Bugs in the AT&T Toolchest program 'nmake'
Summary: there's at least one
Keywords: make, nmake, state files
Message-ID: <9391@alice.UUCP>
Date: 25 May 89 05:44:47 GMT
References: <1640@internal.Apple.COM> <6561@ardent.UUCP> <17678@mimsy.UUCP>
Organization: AT&T Bell Laboratories, Murray Hill NJ
Lines: 20

In article <17678@mimsy.UUCP>, dbk@mimsy.UUCP (Dan Kozak) writes:
> This may only show my naivete, but has anyone implemented a make that
> uses the actual contents (i.e. a CRC or other checkvalue) to compute
> whether something should be re-made?  In fact, you could first send
> the file thru one of the many "de-commenting" programs that were
> floating around comp.lang.c a few months ago, if you didn't want a
> file recompiled when you just changed the comments. (and some people
> thought those programs were useless :-)


mk can support this (but it costs you cycles).
the syntax is (mk sticks per rule attributes between double colons)

goo:Pcmp:	foo
	put your commands here

the semantics are that when mk wants to know if goo is out of date w.r.t.
foo, it sh -c's whatever string comes after P and terminated by the traling :.
a non-zero exit status means it is out-of-date. so in the above example,
goo is up to date if it is the same as foo, independent of any datestamps.