Subject: Re: CVS commit: pkgsrc/mk
To: Julio M. Merino Vidal <>
From: Robert Elz <kre@munnari.OZ.AU>
List: pkgsrc-changes
Date: 08/02/2003 21:43:42
    Date:        Sat, 2 Aug 2003 16:09:36 +0200
    From:        "Julio M. Merino Vidal" <>
    Message-ID:  <>

  | Or simply
  | use the .extract_done cookie file instead of creating another one.  In some
  | situations you will have the message repeated (doing make fetch and make
  | extract manually), but IMO that's acceptable.  Don't you think so? ;)

That sounds like a reasonable compromise to me.   However, I'd still
really like a variable (so it can be set on the command line) to avoid
doing this stuff - with that, doing make fetch, make checksum, make
extract manually need not necessarily generate the message multiple
times (if the user doesn't want it to).

One thing to be wary of in having the message produced early (before the
fetch completes), is that it advises ^C to abort (and says to be sure to
"make clean").   That's fine for later aborts, but if the ^C happens during
the fetch, a partial distfile will be left, which "make clean" doesn't
remove (and a very good thing it doesn't), but which will prevent the
fetch from doing anything the next time around (instead the build will
fail with checksum errors).

This alone may make it a better idea to print the message at the extract
(pre-extract) phase, and simply suffer the few (very few) packages where
choices can affect what is fetched (the worst that cases is some unnecessary
fetching, nothing should actually fail because of it - even if the
unnecessary fetching is something extreme like netscape7 for the wrong