Milter-greylist

Hi,

I've made a few modifications in order to make milter-greylist perform
more responsive and faster at bigger loads:

      * Bucketed search of autowhitelisted and pending entries rather
        than write-locked only-one-at-the-time full list search
      * Large buffer for writing the dumpfile (10M)
      * New option line for the greylist.conf that allows you to disable
        the comment at the end of each dumpfile line. Apparently
        formating the date and time is quite time/cpu consumable task -
        especially if you have to do it a few hundred thousand times in
        a row. The config file option is 'dump_no_time_translation'.

...and a fix to what I consider to be a bug:
      * 'mxsync' client connection is set non-blocking, but there were
        no checks if fprintf() actually delivers the complete sync
        message to the peer or just the part of it.
      * Also, when reading the 'mxsync' client connection, being non-
        blocking, fgets() can bail out with NULL and EAGAIN which is not
        fatal error for the connection.

Best regards,

Ranko

Use this one instead!

Previous patch introduces bug in textdump of autowhitelisted entries and
would cause the loss of autowhitelisted entries...

R.

On Fri, 2006-01-06 at 01:15 +0200, Ranko Zivojnovic wrote:
> Hi,
> 
> I've made a few modifications in order to make milter-greylist perform
> more responsive and faster at bigger loads:
> 
>       * Bucketed search of autowhitelisted and pending entries rather
>         than write-locked only-one-at-the-time full list search
>       * Large buffer for writing the dumpfile (10M)
>       * New option line for the greylist.conf that allows you to
> disable
>         the comment at the end of each dumpfile line. Apparently
>         formating the date and time is quite time/cpu consumable task
> -
>         especially if you have to do it a few hundred thousand times
> in
>         a row. The config file option is 'dump_no_time_translation'.
> 
> ...and a fix to what I consider to be a bug:
>       * 'mxsync' client connection is set non-blocking, but there were
>         no checks if fprintf() actually delivers the complete sync
>         message to the peer or just the part of it.
>       * Also, when reading the 'mxsync' client connection, being non-
>         blocking, fgets() can bail out with NULL and EAGAIN which is
> not
>         fatal error for the connection.
> 
> Best regards,
> 
> Ranko
> 
> 
> 
> ______________________________________________________________________
> YAHOO! GROUPS LINKS
> 
>      1.  Visit your group "milter-greylist" on the web.
>           
>      2.  To unsubscribe from this group, send an email to:
>          milter-greylist-unsubscribe@yahoogroups.com
>           
>      3.  Your use of Yahoo! Groups is subject to the Yahoo! Terms of
>         Service.
> 
> 
> ______________________________________________________________________
> 
-- 
Ranko Zivojnovic,
IT Director/CTO          ranko@...

Spidernet Services Ltd., Tel: +357 22 844844
Nicosia, Cyprus          FAX: +357 22 669470

Too much coffee and not enough sleep I guess...

You also need this small patch to complete parsing of the
'dump_no_time_translation' option.

Best regards,

Ranko

On Fri, 2006-01-06 at 12:26 +0200, Ranko Zivojnovic wrote:
> Use this one instead!
> 
> Previous patch introduces bug in textdump of autowhitelisted entries
> and
> would cause the loss of autowhitelisted entries...
> 
> R.
> 
> On Fri, 2006-01-06 at 01:15 +0200, Ranko Zivojnovic wrote:
> > Hi,
> > 
> > I've made a few modifications in order to make milter-greylist
> perform
> > more responsive and faster at bigger loads:
> > 
> >       * Bucketed search of autowhitelisted and pending entries
> rather
> >         than write-locked only-one-at-the-time full list search
> >       * Large buffer for writing the dumpfile (10M)
> >       * New option line for the greylist.conf that allows you to
> > disable
> >         the comment at the end of each dumpfile line. Apparently
> >         formating the date and time is quite time/cpu consumable
> task
> > -
> >         especially if you have to do it a few hundred thousand times
> > in
> >         a row. The config file option is 'dump_no_time_translation'.
> > 
> > ...and a fix to what I consider to be a bug:
> >       * 'mxsync' client connection is set non-blocking, but there
> were
> >         no checks if fprintf() actually delivers the complete sync
> >         message to the peer or just the part of it.
> >       * Also, when reading the 'mxsync' client connection, being
> non-
> >         blocking, fgets() can bail out with NULL and EAGAIN which is
> > not
> >         fatal error for the connection.
> > 
> > Best regards,
> > 
> > Ranko
> > 
> > 
> > 
> >
> ______________________________________________________________________
> > YAHOO! GROUPS LINKS
> > 
> >      1.  Visit your group "milter-greylist" on the web.
> >           
> >      2.  To unsubscribe from this group, send an email to:
> >          milter-greylist-unsubscribe@yahoogroups.com
> >           
> >      3.  Your use of Yahoo! Groups is subject to the Yahoo! Terms of
> >         Service.
> > 
> > 
> >
> ______________________________________________________________________
> > 
> 
> ______________________________________________________________________
> 
> YAHOO! GROUPS LINKS
> 
>      1.  Visit your group "milter-greylist" on the web.
>           
>      2.  To unsubscribe from this group, send an email to:
>          milter-greylist-unsubscribe@yahoogroups.com
>           
>      3.  Your use of Yahoo! Groups is subject to the Yahoo! Terms of
>         Service. 
> 
> 
> ______________________________________________________________________
>

Ranko Zivojnovic <ranko@...> wrote:

> You also need this small patch to complete parsing of the
> 'dump_no_time_translation' option.

Could you repost a full patch integrating that one? 

-- 
Emmanuel Dreyfus
Publicité subliminale: achetez ce livre!
http://www.eyrolles.com/Informatique/Livre/9782212114638/livre-bsd.php
manu@...

On Fri, 2006-01-06 at 18:57 +0100, manu@... wrote:
> Ranko Zivojnovic <ranko@...> wrote:
> 
> > You also need this small patch to complete parsing of the
> > 'dump_no_time_translation' option.
> 
> Could you repost a full patch integrating that one? 

Here it is.

Best regards,

R.

Ranko Zivojnovic <ranko@...> wrote:

> Here it is.

I've integrated this for the next release. One comment on style for your
next contribution: try to stick to the existing style, that makes the
source easier to read.

- No line longer than 80 chars
- space after if

-- 
Emmanuel Dreyfus
http://hcpnet.free.fr/pubz
manu@...