[ On Sun, November 29, 1998 at 18:37:39 (+1100), John Birrell wrote: ]
> Subject: Re: kernel source
>
> Greg A. Woods wrote:
> > [ On Sat, November 28, 1998 at 18:43:48 (-0800), Jason Thorpe wrote: ]
> > > Subject: Re: kernel source 
> > >
> > > You really want the snapshot building to provide the sources he/she used.
> > 
> > Indeed.  It really shouldn't be all that hard to tag the repository from
> > the working directory used by the snapshot build and then "cvs export
> > -kv -rSNAP_TAG" the sources based on that tag to create the canonical
> > source distribution for a given tagged "release".
> 
> You want each portmaster to tag the whole repository for each snapshot?
> That would be yet another reason for NetBSD not to release snapshots
> very often. A simple date checkout makes more sense. Consider a daily
> snapshot (on those architectures where you actually have time to
> compile the whole tree in a day 8-).

That would be OK so long as the build is done from a fresh working
directory checked out using the intended date.  With something as big as
the NetBSD tree, and esp. running over a WAN, that's the only way to
ensure you don't have any version skew without tagging.

However there are other advantages to tags, such as seeing which
revision of a file was used by referencing only the repository.

Also, the person building the snapshot doesn't need an extra copy of the
tree if they use a tag (except to build the source tar, but with a tag
the tar can be built on any machine).

I don't see what the problem with tagging is.  You simply start the tag
running in a separate window, after you've done the build, minimize it,
and go on to doing something else.  It's called "multitasking"!  ;-)

(the only thing they can't do while tagging happens is to run any cvs
command that would change the state of their working directory)

-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods@acm.org>      <robohack!woods>
Planix, Inc. <woods@planix.com>; Secrets of the Weird <woods@weird.com>