[Teammetrics-discuss] Problem in commitstat

[Andreas Tille]

On Thu, Aug 11, 2011 at 07:54:30AM +0530, Sukhbir Singh wrote:
> > 1. Get healthy again first!
> 
> Thank you, I am much better after sleeping peacefully :)

:-)
 
> > 2. Commit whatever you did for the mbox translation for listmaster
> >  I might have a look myself if needed, but we should bring this
> >  foreward.
> 
> Two possible approaches on how this can be done and I will commit the
> code today soon.

OK.  As I sais: Just commit it to our repository and I'll start testing.

> > $ ./commitstat.py
> > Traceback (most recent call last):
> > File "./commitstat.py", line 222, in <module>
> 
> It's good that you are testing this! Let me know the results. But be
> careful about committing the password :)
> 
> I have a solution for this, which is by far the easiest. How about to
> decrypt the private key, we require that you enter the password when
> the script runs? This IMHO should take care of the problem of saving
> the password.

Well, in my normal workflow when I'm sshing to several hosts I'm using
ssh-agent which lets me type my password only once per session.  It
would be really great if python-paramiko would be able to work together
with ssh-agent somehow.  This would enable us to not need to handle any
password at all.  We also could give our user name on alioth as command
line argument which should enable us using the script unchanged as it
is in git.

I think I managed to work with ssh-agent formerly - but currently it
does not work.  Could you try this approach?
 
> Also, we should not use either yours or my key :) We can create some
> other account for this purpose. How does that sound?

If I remember correctly alioth admins were not very comfortable with
creating such a fake user account when we asked previousely.

> > Please note that I added message_id to listspam table and commit_id
> > to commit table.  The code needs to be adapted ...
> 
> Done for listspam table, but what is `commit_id`?

git log tells you about the ID of the commit and svn has the revision
numbers (svn log | grep ^r) which could serve as ID (together with the
project it can be used as primary key - see my last commit).  We simply
need a way to prevent us from double/multiple entries in the database.
 
> As far as the NNTPstat is concerned, Gmane disconnects in between,
> which is weird. I think best would be to implement the approach you
> suggested but first I will take care of the mbox filter today.

Yes, the mbox filter should gain priority over the other tasks.

Kind regards

      Andreas.

-- 
http://fam-tille.de