[tpop3d-discuss]mailbox->apply_changes() at connection_shutdown().

[Chris Lightfoot]

On Fri, Sep 30, 2005 at 06:48:55PM +0200, Jacek D?bowczyk wrote:
> Hi,
> 
> probably I found bug in use of mailbox->apply_changes(). This is called only
> when when client sends QUIT command. Therefore in situation when client
> sends number of DELE and simly closes connection, these commands won't be
> applied (in my opinion they should).

um. That's a requirement of the RFC, in fact -- you only
go into the UPDATE state when a QUIT command is correctly
issued. Actually in the CVS version there's code to do
much as you describe, with an option to switch this off
(no-commit-on-early-close).

> I'm attaching patches, which repairs this bug. Additionaly I'm using
> cucipop-style logging in my system, which is placed in these patches.

ok -- I will look at adding this.

> --- netloop.c.orig	Sun Jul 31 18:36:40 2005
> +++ netloop.c	Sun Jul 31 18:36:40 2005
> @@ -462,8 +462,12 @@
>          if (c->cstate == closed) {
>              /* We should now log the closure of the connection and ending
>               * of any authenticated session. */
> -            if (c->a)
> +            if (c->a) {
>                  log_print(LOG_INFO, _("connections_post_select: client %s: finished session for `%s' with %s"), c->idstr, c->a->user, c->a->auth);
> +                if (c->m)
> +                    log_print(LOG_NOTICE, _("%s[%d] %s%s %d, %d (%d), %d (%d)"), c->a->user, c->a->uid, c->remote_ip, (c->secured ? ":SSL" : ""),
> +                              time(NULL) - c->start_time, c->m->numdeleted, c->m->sizedeleted, c->m->num, c->m->totalsize);
> +            }
>              log_print(LOG_INFO, _("connections_post_select: client %s: disconnected; %d/%d bytes read/written"), c->idstr, c->nrd, c->nwr);
>  
>  /*            remove_connection(c);*/



-- 
``One thousand years ago we wouldn't have known the news outside our village.
  I wish things were like that now.''
  (http://themanwhofellasleep.com/gossip2.html)