Re: blends-dev, gsoc 2013

Hello Andreas

On Thu, May 30, 2013 at 10:28 AM, Andreas Tille <andreas@an3as.eu> wrote:

> Also I can assume that for versioning I am going
> to use git.

Yes. Feel free to decide whether you want to start from scratch with a
say blends-gsoc repository or simply use a branch in the existing blends
repository.

A word of warning: Please consider myself as a Git
beginner. It might happen that I will relay on asking you about help in
Git details.

Ok no problem, I am not an expert myself but I think we won't have any problems in this part.

> Do you suggest any other sources for me to check? Any other sources that
> will provide a more solid understanding of how thing are working, if not I
> will search around debian sources to find more info on task files etc

IMHO everything you currently need to study is

git://git.debian.org/git/blends/blends.git

(which also contains the source of the doc at the URL above).

Yes you are right, I also checkout the source of some blends like med-bio and debichem to get some better examples of task and control files.

> (also
> it is a very good opportunity for me now to study about the debtags).

Yes. I would love if you could add some method to merge in DebTags into
Blends techniques. DebTags are in UDD as well. BTW, to help solving bug
#703402 I just wrote an importer for Blends metadata.

The easiest way to
learn about those tables would probably be to

$ psql udd
udd=# \d blends_<tab>

You should get four tables where blends_prospectivepackages is quite
irrelevant for the GSoC task because this are package that are not (yet)
inside Debian and thus do not belong into the Recommends. The other
tables are relevant metadata (which might need to be enhanced in the
GSoC project).

Yes I checked the tables and your script(it was very useful).

> How do you prefer to communicate during the program? We can talk through
> skype,

No way. I will install only software where I have the source for.
Sorry, if this sounds to harsh / stubborn - but we have other ways of
communication.

I agree with you, I just mention it cause a lot of people use it, anyway I also prefer other ways of communication

> irc,

I'm rarely doing IRC but it is fine for me outside work hours if you
ping me in advance to fire up my IRC client.

Ok I will keep that in mind.

> gtalk,

Fine for things that need to be sorted out quickly.

> debian mailing list etc.

This is my prefered way because it caries sustainable information that
is properly archived and available for everybody who might be
interested.

Ok that is also fine from me

No questions so far. I personally would start by creating SQL queries
(functions if needed) to assemble the needed information. Once this is
there it could be wrapped up in a Python script that writes the task
file. Keep it simple for the first shot - we will enhance it later.

So you mean for the begging write a python script which will generate the task files for a blend using information from udd?(sorry for maybe repeating the above sentence but I got a little confused)

The blend_dependencies table contains the package dependencies for the blends' latest release?(because I do not see a column containing the version of a blend)

Also in the blend_dependencies table, what do the four possible values (d, i, r, s) of the column dependency mean?

Kind regards

Emmanouil