[Teammetrics-discuss] Commitstat using key authentication?

[Andreas Tille]

On Mon, Sep 05, 2011 at 05:26:17PM +0530, Sukhbir Singh wrote:
> > The problem is that in the past I observed a lot of encoding errors with
> > perfectly valid mails and names.  It just happens that names contain
> > non-ASCII characters and the mails are in different encodings than
> 
> That is true and I have observed such messages, but ..
> 
> > UTF-8.  I would like you to check the logfiles for such cases carefully
> > because I'm pretty sure we are missing a certain amount of valid mails
> 
> I tried fixing these errors but if Python fails to recognize the
> encoding, there is not much we can do :)

Well, I think that every problem of this kind is solvable to some
extend.  The fact that Python fails to parse a string as UTF-8 is no
reason not to try other means.  I'm not an expert in this but somehow
any valid e-mail can be read  - and so I see no reason not to read it
using Python as well.  If you (and I as well admittedly) have no idea
how to do this you need to seek external help in relevant mailing list /
forums.  Just strip down the message to a possibly small mbox (keeping
just those fields which are needed, strip the content to a basic minimum
to not bore the people you are asking for help) and check whether this
minimal mbox shows the problem you can not solve.  Then put it at some
place where people can have a look and provide a short code snippet
which reproduces the problem reading the mbox.

Once I did so I usually got reasonable responses which helped solving
the problem.
 
> > IMHO the best way to find this out is to pick about 20 specific messages
> > in different mboxes for each type of problem and inspect the mail
> > carefully.
> 
> Ok, I am on it.

I also wonder whether we should add these mails to listspam with the
reason "cannot parse" or something like this.  Once we might inspect the
database we might see some common pattern.
 
Kind regards

       Andreas.

-- 
http://fam-tille.de