neal@denelcor.UUCP (Neal Weidenhofer) (03/07/84)
**************************************************************************
A couple of questions/gripes about readnews and associated topics:
1. Why does "U" (unsubscribe) take so long -- sometimes
> 25 min (on a moderately loaded pdp-11/44)?
2. I am trying to split my news reading into two parts --
"work" and "personal". To do this, I created a couple
shell scripts that look like:
cp x .newsrc
readnews
cp .newsrc x
The problem is that the second cp USUALLY doesn't work --
after I exit readnews and get my prompt back, x is
unchanged. About one time in ten, x is updated properly.
I can (and do) always redo the cp by hand and it works
then. Can anyone help?
Regards,
Neal Weidenhofer
Denelcor, Inc.
<hao|csu-cs|brl-bmd>!denelcor!nealdmmartindale@watcgl.UUCP (Dave Martindale) (03/08/84)
The U command itself takes almost no time. But finding the first article in the next group may take a long time indeed if it is a group you haven't been reading regularly. Readnews looks for articles in ascending sequence one at a time. If the next group is net.unix-wizards and the first unexpired article on your machine is number 5000, and you haven't read anything in this group before, it will look for article 1, then 2, then 3, until it finally gets to 5000. You may wait a long time. A quick-and-dirty fix is to grep unix-wizards in /usr/lib/news/active; the number is the highest-received article to date; say it's 6000. Then find the unix-wizards line in your .newsrc and alter it to read net.unix-wizards: 1-5900 Ugh. I believe someone posted some code to deal with this, but we don't have it installed here either. A reasonable way to handle this is to have readnews read the directory, build a bit map of articles which do exist, invert the map to become a set of articles which don't exist, and then or this into the bit map obtained from the user's .newsrc so that all the non-existent articles appear to have been read. I don't have time to implement it though. Dave Martindale
dave@utcsrgv.UUCP (Dave Sherman) (03/08/84)
~| From: neal@denelcor.UUCP (Neal Weidenhofer)
~| 1. Why does "U" (unsubscribe) take so long -- sometimes
~| > 25 min (on a moderately loaded pdp-11/44)?
It doesn't. I suspect what's happening is that after the U, you're moving
on to a newsgroup where, for one reason or another (e.g., new user,
faulty .newsrc), readnews is trying to start at a low numbered
article while the next valid article is high-numbered (in a busy
newsgroup). It tries to open every consecutively numbered file until
it finds one which works. Hence the delay. An easy fix (for you) is to
change your .newsrc. There has also been a fix posted to the net.
~| 2. I am trying to split my news reading into two parts --
~| "work" and "personal". To do this, I created a couple
~| shell scripts that look like:
~| cp x .newsrc
~| readnews
~| cp .newsrc x
~| The problem is that the second cp USUALLY doesn't work --
~| after I exit readnews and get my prompt back, x is
~| unchanged. About one time in ten, x is updated properly.
There's a much better way. I use it so my wife can read a few
newsgroups. Create a file in your bin (call it "pnews" for
personal news if you like) which contains:
echo Personal news coming up...
HOME=/u4/dave/pnews
NEWSRC=/u4/dave/pnews/.newsrc
NEWSBOX=/u4/dave/pnews
NAME="<if you want a different name for postings>"
export HOME NAME NEWSBOX NEWSRC
readnews
Then create a subdirectory pnews. Presto - you have a directory
in which you can run separate news. As an added bonus, all saved
articles will go into that directory, so you can isolate your personal
from work-related files. Now run pnews, and unsubscribe to all
work-related groups. In your regular readnews, unsubscribe to
all non-work-related groups.
Hope this helps. Let me know if it works for you.
Dave Sherman
Toronto
--
{allegra,cornell,decvax,ihnp4,linus,utzoo}!utcsrgv!daverees@apollo.uucp (Jim Rees) (03/19/84)
Here is the version of getnextart(), in readr.c, for news 2.10 or
later, to read the directory instead of trying to open each article.
This will make readnews faster for new users. I don't remember now
who wrote this originally. It had a bug or two having to do with
newsgroups with 0 articles in them, which I have fixed.
/*
* Find the next article we want to consider, if we're done with
* the last one, and show the header.
*/
getnextart(minus)
int minus;
{
int noaccess = 0;
struct direct dir;
long nextnum, tnum;
long atol();
if (minus)
goto nextart2; /* Kludge for "-" command. */
if (bit == obit) /* Return if still on same article as last time */
return 0;
sigtrap = FALSE;
nextart:
dgest = 0;
if (bit < 1 && !rflag)
bit = 1;
/* If done with this newsgroup, find the next one. */
while (ngsize <= 0 || ((long) bit > ngsize) || (rflag && bit < 1)) {
int i;
if (i=nextng()) {
if (actdirect == BACKWARD) {
fprintf(ofp, "Can't back up.\n");
actdirect = FORWARD;
continue;
}
else if (rfq++ || pflag || cflag)
return 1;
}
if (rflag)
bit = ngsize + 1L;
else
bit = -1;
if (uflag) {
long now;
time(&now);
if (now - timelastsaved > 5*60 /* 5 minutes */) {
printf("[Saving .newsrc]\n");
fflush(stdout);
writeoutrc();
timelastsaved = now;
}
}
}
nextart2:
#ifdef DEBUG
fprintf(stderr, "article: %s/%d\n", groupdir, bit);
#endif
if (rcreadok)
rcreadok = 2; /* have seen >= 1 article */
sprintf(filename, "%s/%d", dirname(groupdir), bit);
if (rfq && goodone[0])
strcpy(filename, goodone);
if (sigtrap) {
if (sigtrap == SIGHUP)
return 1;
if (!rcreadok)
xxit(0);
fprintf(ofp, "Abort (n)? ");
fflush(ofp);
gets(bfr);
if (*bfr == 'y' || *bfr == 'Y')
xxit(0);
sigtrap = FALSE;
}
#ifdef DEBUG
fprintf(stderr, "filename = '%s'\n", filename);
#endif
/* Decide if we want to show this article. */
if (bit <= 0 || access(filename, 4)) {
#ifdef apollo
/* Make sure we can still get at the spool directory */
struct stat stbuf;
if (stat(SPOOL, &stbuf) < 0) {
fprintf(ofp, "Net failure has made news temporarily unavailable.\n");
fprintf(ofp, "Do you want to quit (q) or try again (<RET>)? ");
fflush(ofp);
gets(bfr);
if (*bfr == 'q' || *bfr == 'Q') {
writeoutrc();
xxit(0);
} else
goto nextart2;
}
#endif
/* since there can be holes in legal article numbers, */
/* we wait till we hit 5 consecutive bad articles */
/* before we haul off and scan the directory */
if (++noaccess < 5)
goto badart;
noaccess = 0;
fp = fopen(dirname(groupdir), "r");
if (fp == NULL) {
#ifdef DEBUG
fprintf(stderr, "can't open groupdir (%s)\n", dirname(groupdir));
#endif
goto badart;
}
nextnum = rflag ? 0 : ngsize;
while (fread(&dir, sizeof(dir), 1, fp) == 1) {
if (!dir.d_ino)
continue;
tnum = atol(dir.d_name);
#ifdef DEBUG
fprintf(stderr, "art %s (%ld) next %ld\n", dir.d_name, tnum, nextnum);
#endif DEBUG
if (tnum <= 0)
continue;
if (rflag ? (tnum > nextnum && tnum < bit)
: (tnum < nextnum && tnum > bit))
nextnum = tnum;
}
if (rflag ? (nextnum >= bit) : (nextnum <= bit))
goto badart;
do {
clear(bit);
nextbit();
} while (rflag ? (nextnum < bit) : (nextnum > bit));
obit = -1;
abs = FALSE;
fclose(fp);
goto nextart;
} else
noaccess = 0;
if (ignorenews
|| ((fp = fopen(filename, "r")) == NULL)
|| (hread(&h, fp, TRUE) == NULL)
|| (!rfq && !select(&h, abs))) {
badart:
#ifdef DEBUG
fprintf(stderr, "Bad article '%s'\n", filename);
#endif
if (fp != NULL) {
fclose(fp);
fp = NULL;
}
clear(bit);
obit = -1;
nextbit();
abs = FALSE;
goto nextart;
}
abs = FALSE;
actdirect = FORWARD;
news = TRUE;
hdr();
if ((cflag && !lflag && !eflag) || pflag)
tprint(fp, ofp, FALSE);
if (cflag || lflag || eflag || pflag) {
itsbeenseen(h.ident);
sigtrap = FALSE;
fclose(fp);
fp = NULL;
}
obit = bit;
return 0;
}hansen@pegasus.UUCP (Tony L. Hansen) (03/30/84)
When you add in the new getnextart() routine to readr.c, it may or may not be necessary to also add #include "ndir.h" in front of it. I had to on my site to get it to work right. This may have been obvious to many people, but I thought I'd mention it just in case it wasn't. Tony Hansen pegasus!hansen
neal@denelcor.UUCP (Neal Weidenhofer) (04/06/84)
************************************************************************** Many thanks to all of those who offered help and suggestions with the problems I had with readnews. Here is what I found out. 1. Unsubscribe was slow because I was working with a clean .newsrc file. Whenever our version of readnews goes to a group that isn't mentioned in .newsrc, it tries to open (or stat--I heard both and I haven't had the time to check) each file by number until it finds one that exists. As it turned out, the group I was trying to unsubscribe which took 25 minutes was net.jobs. Net.jobs is followed by net.jokes with somewhat more than 5000 expired entries. Now I know how long it takes to do 5000 unsuccessful opens (or stats). A fix for this problem has been posted (not by me). 2. The reason the "cp .newsrc x" was failing in the shell script immediately after readnews is apparently the fact that readnews forks a new process to update .newsrc and exits so that the terminal is freed sooner. In any case, the NEWSRC environment variable did the trick for me. Ln would probably have worked also but this is cleaner. Thanks again and I apologize for taking so long to answer. Regards, Neal Weidenhofer Denelcor, Inc. <hao|csu-cs|brl-bmd>!denelcor!neal