On Sun, Mar 29, 2009 at 02:05:04PM -0400, Perry E. Metzger wrote: > > Alistair Crooks <a...@pkgsrc.org> writes: > > This is veering off topic, I fear, but a datapoint: > > > > We used a yyyymmdd-based time value for version numbering of the > > pkg_install tools, derived from embedded RCS Ids, for years (to check > > tools had sufficient features for what they were going to be asked to > > do). This was across NetBSD and pkgsrc branches. > > That's fine, but those are > > a) essentially monotonic because of your management practices > b) not based on build time (which is my main complaint about things like > the old make(1) version stamp. > > Also, the pkg tools are under quite active development and have > substantial interface shifts with time, and you need to automatically > decide when to rebuild them so you *need* a version.
If you want to check for features, then I suggest you add a -Q: argument to make(1), to be used for querying for features. A non-existent -Q argument would mean the feature wasn't available. Other uses would be: % make -Q repeatable-builds repeatable-builds supported % echo $? 0 % make -Q me-a-crocodile-sandwich-and-make-it-snappy me-a-crocodile-sanewich-and-make-it-snappy not supported % echo $? 1 i.e. de-couple the feature from the date. The two are obviously not related in this case. Regards, Alistair
