REVU: Automated Package Checks

[Nathan Handler]

On Fri, Jan 23, 2009 at 7:10 PM, Kjeldgaard Morten <mok at bioxray.au.dk> wrote:
> The point is, it doesn't solve our problem, because at some point, a
> human being needs to have a look at the package. Even if all packages
> were perfect, we still could not handle one review plus an upload with
> the current activity of MOTUs.
>
> In addition, having this automatic checking would not change the rate
> of arrival of new packages. Our rate of processing is less than the
> rate of arriving packages, and consequently the pool of packages
> awaiting review is constantly increasing.

I can not speak for other MOTUs, but when I review a package on REVU,
I normally subscribe to it. That way, I can review it again once a new
version is uploaded. If these uploaders abandon the package, I usually
move on and review a new package on REVU. Hoever, if the uploader is
active, and prepares new versions as-needed, I am usually busy
re-reviewing that package, so I don't review any new ones. For me, I
would much rather keep working on one package and get it to the point
where it can be uploaded than review a lot of packages only once.

>
> A second point is, that no matter how sophisticated the program, it
> will not be able to solve issues that are common with many of the
> packages. F.ex. that the package has to be split into several
> subpackages, which often is a point of discussion, that the
> description is not understandable, that files need to be removed from
> upstreams tarball, etc. etc. There are lots and lots of issues that
> could never be detected automatically.

You are correct. We would only be able to test for some of the more
basic stuff. However, this will reduce the number of times that the
contributor would need to upload, wait for a review, and revise. The
fewer times that they need to do that, the faster we can get the
package uploaded and out of REVU.
>
> Thirdly, and most importantly, is the personal interaction we get with
> the uploaders, and in this regard the simple things people are asked
> to fix is often a useful beginning. It gives you an opportunity to
> judge the qualifications and personal qualities of the uploader, and
> it tells you if the uploader is truly interested in doing some work.

REVU is not designed to form relationships. If the MOTU is interested
in getting to know the uploader, they should initiate a conversation
either on IRC or via email. Also, these automated checks will still
allow the MOTU to see if the uploader is really interested in working
on the package. If they are willing to take the time to make the
corrections proposed by the checks, they are probably also willing to
make any corrections proposed by the MOTU.
>
> So, in my opinion, REVU is a very good tool already that fulfills it's
> purpose very well. What is lacking is the involvement of more MOTUs.

Yes, REVU is a great tool. And yes, having more MOTUs reviewing
packages would solve a lot of the problems. However, we can not force
people to spend time on REVU. As a result, we need to do what we can
to make REVU as easy to use and efficient as possible.