Path: utzoo!utgpu!watserv1!watmath!att!pacbell.com!ames!think.com!mintaka!bloom-beacon!eru!hagbard!sunic!dkuug!rcbal!fk
From: fk@rci.dk (Flemming Kraglund)
Newsgroups: news.software.b
Subject: Re: sites changing a message id
Summary: cnews speedup has bug in tear
Message-ID: <1991Jan23.095309.2952@rci.dk>
Date: 23 Jan 91 09:53:09 GMT
References: <CQP+%BA@b-tech.uucp> <1991Jan12.234625.17232@zoo.toronto.edu>
Organization: RC International, Copenhagen, Denmark
Lines: 86

henry@zoo.toronto.edu (Henry Spencer) writes:

>In article <CQP+%BA@b-tech.uucp> zeeff@b-tech.ann-arbor.mi.us (Jon Zeeff) writes:
>>Apparently various sites decided to change/add the message id on this
>>article making it look like it had been posted many times.  Sounds
>>like antisocial behavior on the part of some software somewhere.

>Last time we ran into this, the diagnosis was messed-up headers in the
>original (which can happen as a result of using an old version of C News
>and posting an article whose early lines begin with white space, although
>the odd format of the second Message-ID line argues weakly against this)
>which caused B News not to see the Message-ID header... whereupon it adds
>one with a message-ID of its own invention.  So the single messed-up
>original is rebroadcasted, under a new message-ID, by every B News site 
>it hits.  Ick.

>I'm not sure this entirely accounts for this case, since the new ID looks
>much more like a C News ID.  That's curious, because C News definitely does
>*not* make up a message-ID to fill the gap.  Old C Newses just passed on
>the article unmodified; new ones drop it on the floor.

I had that problem too, the Message-ID being put into the body and not
into the header. I am running Brendan Kehoe's extensions and found that
there is a bug in the c-version of tear making it belive that if the
first line started with a space or tab it was a continuation header
line, so it is possible that if a site is running with Brendan's
extensions and is recieving/storing articles via inews we get the
mentioned senario.

I have been trying to contact brendan the last 2 weeks but I get those
can not send for one week messages back. So going behind Brendans back
I will include a patch for tear.c.
While I am at it I will include another one for anne.misc.c which may
dump core on some ocations.
(Brendan has confirmed that he has recived the bug report on this one)

b.r.   Flemming Kraglund  fk@rci.dk

Patches follows

*** tear.c-	Thu Dec  6 17:51:39 1990
--- tear.c	Sun Jan  6 16:18:44 1991
***************
*** 48,54
  				if ((!strchr(lp,':'))||
  				    ((curline>1)&&(*lp == '\n')))
  					doit = 1;
! 			}
  			if (doit) {
  				inbody = 1;
  				fclose(fout);

--- 48,55 -----
  				if ((!strchr(lp,':'))||
  				    ((curline>1)&&(*lp == '\n')))
  					doit = 1;
! 			} else if (curline == 1)
! 				doit = 1;
  			if (doit) {
  				inbody = 1;
  				fclose(fout);
*** anne.misc.c-	Mon Dec 17 15:06:03 1990
--- anne.misc.c	Mon Dec 17 15:06:54 1990
***************
*** 46,55
      */
    t = u = safemalloc(strlen(s) + 1);
    
!   do {
!     if (*s != ' ' && *s != '\t')
!       *(t++) = *s;
!   } while (*s++);
    
    *t = '\0';
  #ifdef	DEBUG

--- 46,52 -----
      */
    t = u = safemalloc(strlen(s) + 1);
    
!   for (; *s; ++s) if (*s != ' ' && *s != '\t') *t++ = *s;
    
    *t = '\0';
  #ifdef	DEBUG
-- 
Flemming Kraglund             fk@rci.dk                  This is my OWN opinion