Date: Fri, 22 Aug 1997 09:29:34 PDT From: Bill Fenner <fenner@parc.xerox.com> To: freebsd-ports@freebsd.org Subject: Re: Versioning bsd.port.mk Message-ID: <97Aug22.092938pdt.177486@crevenia.parc.xerox.com>
next in thread | raw e-mail | index | archive | help
Tim Vanderhoek <tim@x22> wrote: >Yes, every port should depend on a specific revision of >bsd.port.mk, not just the complex ones. I disagree; a port that uses only $DISTNAME and $MASTER_SITES will never break (unless we're morons). Satoshi Asami <asami@cs.berkeley.edu> wrote: >The reason why I mentioned putting the variable in the Makefile was >because people are already having a hard enough time remembering to >commit all required files Hmm. That was part of my reluctance to make it a seperate file -- wasn't sure if people would remember it. >By the way, do we really agree that we should force this variable >(modulo manual overrides, of course) to exactly match the timestamp of >bsd.port.mk? No, I thought that it should be "this-version-or-newer", just like when using __FreeBSD_version or BSD defines. >can we use the date in the $Id$ field in the port's Makefile? No, that will cry wolf too much. Plenty of ports will still work, while printing a warning that they might not work, so when a port build doesn't work, the user will probably ignore the warning since it was wrong so many times before. I guess there's no "on error" target in Make -- maybe we could forget all this versioning stuff and print a message about making sure your bsd.port.mk was up to date when a build error occurs. The best I can still think of doing is either a Makefile variable or a files/ file; each has its pros and cons but there's no clear winner in my book. Bill P.S. This is probably my last word on the topic for now, since I'm leaving for a couple of weeks, and presumably when we're back from our honeymoon this will all have been decided and committed.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?97Aug22.092938pdt.177486>