Re: Documentation issues on the new website

[Malcolm Tredinnick <malcolm commsecure com au>]

On Wed, Aug 14, 2002 at 11:44:23AM +1000, Jeff Waugh wrote:
[...]
> > If you can make the hackery work, I think we would be better off with a
> > system that automagically pulls the docs from cvs. My anecdotal experience
> > (based on past efforts to keep the doctable updated) is that a second task
> > beyond checking a doc in will often be forgotten.
> 
> Great, I'll push on with this then.
> 
> Have you guys considered doing validation on CVS checkin? I guess it's kinda
> hard with docs all over the CVS tree...

My experience (from a work project where I had set up some automatic
validation of stuff on checkin) is that it is actually a handicap. From
time to time (more often than I had expected) somebody will want to
check in something that is not quite correct so that another person can
work on it or try to get the bugs sorted out. This instantly triggers an
email or whatever happens on validation failure. It doesn't necessarily
help to send the pseudo-checkin to a branch, either, since the automatic
CVS scripts would then need to check the branch name of the recent
checkin and only go to work if it is on the list of "interesting"
branches (say, HEAD and gnome-2-0 for now).

I would suspect this scenario is even more common when most of the
interaction we have is via IRC or mailing lists and developers are
located around the world.

It is not impossible to do this (have a script that is run on every
single cvs checkin and have the script be very intelligent about when it
does "real" work), but it would probably be better to just periodically
checkout and validate "offline" (i.e. away from the main CVS machine).

Cheers,
Malcolm