Path: utzoo!utgpu!jarvis.csri.toronto.edu!rutgers!usc!zaphod.mps.ohio-state.edu!wuarchive!kuhub.cc.ukans.edu!anu-news!list
From: munnari!anu.anu.oz.au!gih900@CSC.DNET (Geoff Huston)
Newsgroups: news.software.anu-news
Subject: RE: Comments on V5.9C
Message-ID: <9001191517.AA14344@uunet.UU.NET>
Date: 19 Jan 90 22:16:21 GMT
Sender: ANU-NEWS Discussion <ANU-NEWS@NDSUVM1.BITNET>
Reply-To: Geoff Huston <munnari!anu.anu.oz.au!gih900@CSC.DNET>
Lines: 119

>I have just upgraded to V5.9c from V5.8b and I have several comments/problems.
>
>1.	The NNTP batch file changed names again! (From NEWS.BATCH to NNTP.BATCH
>	to NNTP_pid.BATCH.  At first I thought of this as a minor annoyance,
>	remedied by changing the RENAME in my batch job.  But, NOOO, it has
>	caused real problems.  Previously I went through some pains to ensure
>	the ADD FILE file.name.*/DELETE got the files in order as received,
>	with "RENAME NNTP.BATCH.* NNTP.REVBAT.".  That reversed the versions
>	(*.*.high became *.*.1, *.*.-1 -> *.*.2, etc.)  Now, when I do a
>	"RENAME NNTP*.BATCH.* NNTP.REVBAT.", the files end up in an order
>	depending on the NNTP job's PID!  I am getting an annoying number of
>	followups before seeing the original message.
>
>	I will (unless someone else has done it) write and sent out code
>	to make the name something like NNTP_900104_125631_03AB.BATCH.
>	(that's NNTP_yymmdd_hhmmss_pid.BATCH)  Hopefully in a very few days..
     
Here's my solution:...
     
NNTP_SERVER.C
     
                            /* file names  - change this line */
#define NEWSBATCH       "NEWS_MANAGER:NNTP_%s_%X.BATCH"
     
/* new procedure  */
     
static char timestr[14];
     
static
char *time_str()
{
  int ctime;
  struct tm *stm;
     
  time(&ctime);
  stm = localtime(&ctime);
  sprintf(timestr,"%02d%02d%02d%02d%02d%02d",
   stm->tm_year,stm->tm_mon,stm->tm_mday,stm->tm_hour,stm->tm_min,stm->tm_sec);
  return(timestr);
}
   /* ...... */
     
/* char batch[] was previously declared as a local var to add_item */
     
static char batch[FILE_STRLEN] = "";
     
static
add_item(fn,mode,msgbuf,stm)
  char *fn;
  int mode;
  char *msgbuf;
  int stm;
{
  FILE *fpr,
       *fpw;
  char line[BUFF_STRLEN],
       batch[FILE_STRLEN];
  struct stat sbuffer;
     
  if (mode == POST) return(post_add_item(fn,msgbuf,stm));
     
  if (!(fpr = fopen(fn,"r"))) {
    strcpy(msgbuf,"File handler error in NNTP Server.");
    return(0);
    }
  if (!*batch) sprintf(batch,NEWSBATCH,time_str(),getpid());
  if (!stat(batch,&sbuffer)) {
    if (sbuffer.st_size > NEWS_BATCH_SIZE) {
      sprintf(batch,NEWSBATCH,time_str(),getpid());
      fpw = fopen(batch,"w","alq=30","deq=30");
      }
    else if (!(fpw = fopen(batch,"a"))) {
      sprintf(batch,NEWSBATCH,time_str(),getpid());
      fpw = fopen(batch,"w","alq=30","deq=30");
      }
    }
  else {
    sprintf(batch,NEWSBATCH,time_str(),getpid());
    fpw = fopen(batch,"w","alq=30","deq=30");
    }
  if (!fpw) {
    strcpy(msgbuf,"File handler error in NNTP Server.");
    return(0);
    }
  fprintf(fpw,"#! rnews %d\n", (ihave_size > 0 ? ihave_size : 1));
  while (fgets(line,BUFF_STRLEN,fpr)) fputs(line,fpw);
  fclose(fpr);
  fclose(fpw);
  strcpy(msgbuf,"Item successfully spooled.");
  while (!delete(fn));
  return(1);
}
     
that should do the trick - it will be included in the next release.
     
>2.	After OPEN MAIL and deleting an item, I got strange problems.  Once
>	I got something like 'Bad Block Address' messages when I tried to
>	read items after the delete, and the other time the newsgroup
>	directory was totally messed up.
     
any shots at what is going wrong??
     
>3.	Does anyone else have problems with keypad mode being lost after
>	spawning or posting?  I use MG, and when I return to NEWS, my
>	KP7 is 7, etc.  I suspect an SMG$ call may be necessary on return.
>	Comments?
     
I'm interested in any comments as well!
     
>4.	In NNTP_FEED.C, there is a NNTP_MAX_OFFER (??) definition of 500 items
>	to send at one time.  I often will have many more.  Is there a reason
>	it is set so low?
     
To limit the connect time for any single session - make it higher if you are on
a local net, or on a high bandwidth trunk to the NNTP feed node.
     
cheers,
     
Geoff