On Mon, Nov 03, 2003 at 05:28:34PM -0500, Sean Davis wrote:
> On Mon, Nov 03, 2003 at 12:31:16PM -0800, Andrey Petrov wrote:
> > On Mon, Nov 03, 2003 at 03:10:18PM -0500, Sean Davis wrote:
> > > I recently setup my NetBSD/sparc64 machine (Ultra5, 270mhz) with
> > > spamassassin and am running spamd on it. My main i386 machine which actually
> > > receives the mail then (via spamc -d ...) offloads spam processing to the
> > > sparc64: the only annoying part is this.
> > > 
> > > I had a bunch of messages come in at once (which is not unusual) so they
> > > were tossed at spamd, five at a time (the concurrency limit I have on local
> > > delivery so as not to overload the spam processing box), and I saw this as
> > > they were getting processed:
> > > 
> > > (wordwrapped to avoid long lines)
> > > Nov  3 14:55:32 starways /netbsd: hme0: status=21<GOTFRAME,RFIFOVF>
> <snip>
> > 
> > Use hme.c revision 1.39.
> 
> Looking at the diff, it appears all that has changed is the #define HMEDEBUG
> is commented out. The stuff in hme.c::hme_eint() that causes those messages
> isn't wrapped in an #ifdef HMEDEBUG, that I can see. Does the fact that the
> interface doesn't get set into debug mode by default without HMEDEBUG make
> it so that ((status & HME_SEB_STAT_ALL_ERRORS) != 0) (line 991) can never be
> true?
> 
When HMEDEBUG is not defined HME_SEB_STAT_DTIMEXP and HME_SEB_STAT_RFIFOVF
are excluded from HME_SEB_STAT_ALL_ERRORS.

	Andrey