YoSucker Yahoo! mail extractor general discussion list

David Aspinwall | 24 Jan 2004 00:52

Re: messages chopped in half

>>"svref" == David Morse <svref@...> writes:

> My mail flow, if I remember it right:
> yahoo ---> Yosucker ---> localhost:/var/spool/mail/dm ---> pop3d -----> Mozilla

Is this really safe?

From a quick look at the YoSucker code, it seems like it is doing a
flock call on the mail spool file itself (/var/spool/mail/dm in this
case).  I think sendmail actually locks a file with a '.lock' at the
end.  At least that's what the 'movemail' command that comes with
emacs does.  Plus, some versions of sendmail don't use flock, and some
systems don't use sendmail, so getting this right doesn't seem that
easy.  If sendmail is trying to write the file while YoSucker is also
trying to write it, or your mail client or pop server is reading the
file while YoSucker is writing it, mail can get lost or truncated.

I don't tend to trust file locking with sendmail in any case.  What I
do is set PIPE=1 in the config file, and set OUTFILE to a script which
basically looks like:

#!/bin/sh
/usr/bin/spamc -f | /usr/sbin/sendmail -oi myname@...

The spamc is there because I'm using spamassassin; if you aren't, you
could get rid of all up to and including the '|'.

Of course, you should test the sendmail command by itself first to
make sure it works.

(Continue reading)

headers

Vlad Zero | 26 Jan 2004 17:48

Re: messages chopped in half

Well, yeah, as David A. noticed, there are several methods of file lock
protection. There's a more or less standard way of file-locking, the
system 'flock' call that I used in YoSucker, but this solution surely has
a very weak point - it only works if the other applications use the flock
way as well. Many does not, as they employ their own way of file-locking.  
But it would be impossible to take in account all of them. And I don't
think the primary YoSucker user is a heavy traffic production mail server,
but a single person that likes to fetch the its Yahoo messages once in a
while. So I went for the flock, that I considered standard. It surely
brings some risk to the code, but I found it acceptable. But if anyone
wants to write a more complex flocking mechanism, taking in account all
the different ways of flocking, I'll be happy to apply your patch! The
question is right: "Is this really safe?" Not 100%. But there's been
hundreds or maybe even thousands of users since May 1st 2002 (the
first public release) and I never heard of any trouble. 

(David M., you may actually be the first one to experience this kind of
trouble.  Unfortunately. I'll recheck the code, but everything seems to
be fine... Try the solution bellow as well...)

Let me know if the problem persist... Take care,

Vlad
-z3r0-

On Fri, 23 Jan 2004, David Aspinwall wrote:

> >>"svref" == David Morse <svref@...> writes:
>

(Continue reading)