Friday, 8 November 2013

Re: Ubuntu Platform developers BOF session?

Version: GnuPG v1.4.15 (GNU/Linux)

On Nov 08, 2013, at 08:34 PM, Robie Basak wrote:

>I more or less gave up using UDD. The problem is that I can't invest
>exclusively in that workflow. I try, and then sometimes I have a problem
>because of a package import failure. I don't think I touch that many
>packages, yet I've hit the issue several times.

There's no question it can be frustrating if the packages you touch are out of
date, especially because there's nothing you can really do to fix that except
ping Somebody and cross your fingers. Hopefully the session can address
solutions that will empower Ubuntu Developers to fix these problems
themselves when they encounter them.

>So I end up using a different workflow from the beginning. I'm not sure
>what I'd gain in using the UDD workflow if it doesn't work some of the

I touch a lot of packages and the vast majority of them are just fine. I
think the number is around 2% of all source packages failing[*]. When it does
work, UDD has a lot of advantages because you get to work on package updates
using an actual dVCS. So things like merging in new upstreams or new Debian
versions gets really easy, as does iterative development of new package
versions. But there's no question there are some rough edges too, it's just
that I've mostly figured out all of them, so it works pretty well for me.

There's some discussion going on in Debian about implementing the moral
equivalent of UDD using git. As much as I dislike git, if they did it right
and solved some of the trickier problems that are still outstanding for UDD
(e.g. better quilt integration), I'd manage to overcome that and probably
would do all my Debian (non-team) packages using it.

Anyway, I find it more productive to use UDD as the default, and drop down to
apt-get source when the package I need to touch is out of date.


[*] % grep '^Package: ' /var/lib/apt/lists/*_Sources | wc -l

920 / 43047 =~ 2%