Subject: Re: From PR to pkgsrc
To: Todd Vierling <firstname.lastname@example.org>
From: Greg A. Woods <email@example.com>
Date: 12/17/2004 19:05:54
[ On Friday, December 17, 2004 at 11:54:41 (-0500), Todd Vierling wrote: ]
> Subject: Re: From PR to pkgsrc
> You're more than welcome to find us twice to four times the head count of
> *experienced* pkgsrc users to handle pkg-submission PRs, including field
> testing and cleanups. That's what pkgsrc-wip does: it brings the packages
> into a repository where others can do field testing and help clean them up,
> rather than letting them stagnate in a uuencoded tarball.
How many people really use pkgsrc-wip for testing? Do you/we have any
way of even guessing?
Personally I won't touch pkgsrc-wip with a 10-foot pole.
From the use-it-for-testing perspective it's just a useless distraction
to me -- I don't need it just to test other people's experiments and
I've already got more than enough of my own experiments to keep me busy
anyway. Besides if I want to borrow something from it then I can do
that piece-meal without having to actually use it the way it's intended
to be used.
I'd even go so far as to suggest that most other folks can barely keep
up with testing the official pkgsrc-current, let alone also any of the
stuff from pkgsrc-wip. I suppose those who really love the bleeding
edge will appreciate it, but it seems like a lot of effort just for that.
For me it's also much easier to maintain my local changes all together
in one local working directory that's checked out from the official
pkgsrc tree. That way I can follow along with the official developments
by usining "cvs update", and I can also provide patches and updates
against revisions in the official tree, and generally keep my own local
maintenance headaches to a minimum.
I have even taken to following the quarterly branches in order to ensure
my local work is easier to merge with the new stuff from the official
tree in more consistent lumps. I used to try to do my local merges near
the end of each quarterly freeze, but by following the branch I can now
also benefit from whatever maintenance happens to be done on the branch.
Now I can simply run "cvs update -A -r pkgsrc-NEWQ" on a copy of my
local tree and then make the leap into the future with one big jump at a
time instead of trying to take tiny baby steps since more often than not
each step will require rebuilding half the world, no matter how big or
small a step it is, i.e. the fewer steps, the fewer rebuilds. :-)
> In the process, we've actually gained some main pkgsrc folks who "learned by
> doing" in the pkgsrc-wip tree. The pkgsrc-wip effort is actually quite
> successful so far.
I'm not sure I see how that's any different than what might be achieved
by simply being more liberal with commit privileges in the official
tree. I.e. treat the trunk as the work in progress and promote the
quarterly branches as the place users wanting stability need to be.
I do see that pkgsrc-wip may help promote pkgsrc amongst non-NetBSD
users but I'm not sure that it doesn't just spread resources out more
than necessary on the NetBSD front.
> Remember, this *is* a volunteer effort, and few people have non-small
> amounts of time available to devote to it.
Oh, no question about that!
> pkgsrc-wip provides a way to
> distribute that time more effectively.
I do not quite agree with that though. Until and unless "send-pr" is
entirely deprecated for pkgsrc bug reports then I would suggest that
having what amounts to separate trees like that is actually spreading
rare resources too thin.
If the official pkgsrc tree were even just a slight bit more open then
it should be easier to have all available hands focus on one common
"product" and to make the most effective use of GNATS for handling
And no, I'm certainly not asking for commit access to either pkgsrc-wip
nor pkgsrc, nor am I even volunteering to do any more than I already do
via send-pr! I've got too much on my plate already! ;-)
What I am suggesting though is that its very important to have a _very_
easy to use mechanism for third parties (such as myself) to submit bug
reports, change requests, updates, etc., and that there only be ONE
preferred mechanism for such submissions, no matter what it is. (And
I'm implying that pkgsrc-wip is not anywhere near easy enough for
everyone to use for that purpose! :-)
Greg A. Woods
+1 416 218-0098 VE3TCP RoboHack <firstname.lastname@example.org>
Planix, Inc. <email@example.com> Secrets of the Weird <firstname.lastname@example.org>