salkind (07/23/82)
Here is a rather serious bug in newsinput.c: if someone from the net (in B news) replies to an article whose title ends with "- (nf)", the notes system just throws the response away. The problem is that newsinput is expecting the article to be in "almost generic" format (those two lines that start with a # at the beginning of the article). When it doesn't see a # newsinput decides the article is corrupted and throws it away. I propose those two lines needed by notes be moved into the header field of B news. Then normal B news readers won't even see these lines. Newsinput should also be changed to parse the B news format. Then it can even take advantage of the new References: field in B news to build responses instead of base notes automatically. I don't expect this to be that hard; I've already changed newsinput to be able to parse the B news header. I haven't had a chance to test things out yet though; when (and if) I do I will let you all know. We've found several other bugs in the notesfile system (mostly just minor annoyances). Should I report them in this group? I haven't fixed that many. By the way, everyone here who has tried notes seems to be quite pleased with the system. A big improvement over the B news interface. Lou Salkind
essick (07/27/82)
#R:cmcl2:-808200:uiucdcs:23300007:004:842 uiucdcs!essick Jul 27 15:36:00 1982 I just finished installing some changes that seem to take care of the problem of having both a "Re:" prefix and a "- (nf)" suffix in the title of the article. Simply put, if the code has a "- (nf)" suffix, the body will be saved in a scratch file while the "#" is being looked for. If the "#" is not found, the file is rewound, and the code jumps to the portion of the code that parses simple news articles. The tests that I have done on it this afternoon (admittedly few) seem to work fine. I plan on letting it run for a few more days before I send it out to people who would like a copy. The changes that newsinput has gone through since the release are extensive enough that it will probably be easier to post the entire source to that file rather than try to send a diff script. Ray Essick University of Illinois uiucdcs!essick
essick (07/30/82)
#R:cmcl2:-808200:uiucdcs:23300010:004:15212
uiucdcs!essick Jul 30 09:48:00 1982
As promised, here is a version of newsinput that handles
(maybe not perfectly, but a lot closer) the various problems that
were there before. In particular, Malcolm Slaney was kind enough to
provide the code to parse the "Re:" prefixes of news articles. The
problem where articles having both "Re:" and "- (nf)" has also been
taken care of.
To use this new routine, one has to modify "tsearch.c", a copy
of which is in another response. The makefile should also be modified
so that "tsearch.o" is included as a dependency for newsinput and
is loaded with newsinput.
This version seems to be working for me.
Ray Essick
- - - - - - -
#include "structs.h"
#include "globs.h"
#include "newsgate.h"
/*
* newsinput.c
*
* This program will accept an article according the the
* 'A' protocol for news(I). The article will then be inserted in
* the notefile which matches the newsgroup.
* The program will not insert an article into multiple newsgroups!
*
* Modified April 1982:
* The newer version is somewhat more streamlined. It still uses the
* A protocol for intersystem transfer, but since the B news release
* still accepts this stuff, I don't mind. I also am too lazy to bother
* rewriting the header parser to grab the B format. ( someone could
* probably figure a way to simply steal the code out of B news and
* deposit it here, but I have other things I would rather do)
*
* So, the other major difference in the new version is that if the
* title is of the form:
* some title - (nf)
* (or if the suffix is "- (nf)"
* we decide that it has been generated by notesfiles. This means that
* we are allowed to skip text up to a # and then we have a special
* ALMOST generic notesfile format. The header information is
* slightly different.
* see canon.h for the new format examples
*
* Original Coding: Ray Essick Feb 1982
* Modified extensively: Ray Essick April 1982
* Modified to add some neat things
* Malcolm Slaney July 1982
* added code to catch misformed articles
* Ray Essick July 27, 1982
*
*/
main (argc, argv)
char **argv;
{
struct io_f io;
char title[TITLEN + 40]; /* title */
char nf[NNLEN]; /* the notefile */
char fname[WDLEN]; /* hold scratch file name */
char ngroup[100]; /* newsgroup name */
char fromsys[SYSSZ],
origsys[SYSSZ];
char *p,
*q,
*r;
char *suffix;
int count,
atcount, /* count ARPA hosts */
status;
int c;
int i,
j,
stat,
length, /* of title */
notenum;
struct note_f note;
struct note_f note2;
struct auth_f auth;
long newsseq;
struct id_f respid;
struct when_f entered,
whentime;
struct id_f newsid;
char line[WDLEN + 1]; /* scratch */
struct daddr_f where;
FILE * rawnews; /* raw news read from here */
FILE * scr; /* scratch file for holding article */
#include "main.i" /* common init code and such */
rawnews = stdin; /* read from the right file */
/* Parse the header
*/
if (getc (rawnews) != 'A')
{
fprintf (stderr, "Article not protocol A\n");
exit (BAD); /* wrong protocol */
}
fscanf (rawnews, "%[^.].%ld", origsys, &newsseq);
while (getc (rawnews) != '\n'); /* skip to next line */
/* grab the 'newsgroup'
*/
for (i = 0; (i < 99) && ((c = getc (rawnews)) != '\n') && (c != ','); i++)
ngroup[i] = c;
ngroup[++i] = '\0'; /* null terminate */
if (c != '\n')
while (getc (rawnews) != '\n'); /* skip to eol */
/* now the author line
*/
for (i = 0; (i < WDLEN) && ((c = getc (rawnews)) != '\n'); i++)
line[i] = c; /* grab the whole line */
line[i] = '\0'; /* terminate it */
if (c != '\n')
while (getc (rawnews) != '\n'); /* make sure is end */
j = i; /* start at rhs */
while (j && (line[j] != '!') && (line[j] != ':') && (line[j] != '.'))
j--; /* back to sys */
atcount = 0; /* count ARPA hosts parsed */
for (i = 0, j++; (i < NAMESZ); i++)
if ((auth.aname[i] = line[j + i]) == '@' && atcount++)
{
auth.aname[i] = '\0';
break; /* only parse 1 ARPA host */
}
else
if (auth.aname[i] == ' ')
{
auth.aname[i] = '\0'; /* null terminate */
break; /* and get out of here */
}
else
if (auth.aname[i] == 0)
break;
/* build the from system */
for (i = 0; (i < SYSSZ - 1) && line[i] != '!' && line[i] != ':' && line[i] != '.'; i++)
fromsys[i] = line[i];
fromsys[i] = '\0';
/* get the date written
*/
i = 0;
while ((line[i++] = getc (rawnews)) != '\n');
line[i] = '\0'; /* build this line */
getdate (line, &entered); /* and grab it */
/* now the title line
*/
for (i = 0; (i < TITLEN + 39) && ((c = getc (rawnews)) != '\n'); i++)
title[i] = c;
title[i] = '\0'; /* null terminator there */
if (c != '\n')
while (getc (rawnews) != '\n'); /* skip eol */
/* The header is now parsed, all left is text. Check existence
* of the notesfile, permission to write and other things like that
*/
newsgroup (ngroup, nf, NEWSNF); /* alias the bugger */
if (init (&io, nf) < 0) /* well is it there? */
{
printf ("No notesfile: %s\n", nf);
#ifdef NOSUCHWARN
sprintf (line, "Notesfile: %s, newsgroup %s\n", ngroup, nf);
nfcomment (NOSUCHWARN, line, "unknown newsgroup", 0, 0);
#endif
#ifdef NEWNEWS /* save those new ones */
/* this code courtesy of malcolm... */
printf ("Inserting into nfmaint\n");
sprintf (title, "New Newsgroup: %s", nf); /* Blow away old title */
strcpy (nf, NEWNEWS); /* Change the newsgroup */
if (init (&io, nf) < 0)
exit (BAD); /* Give UP */
printf ("Open of %s suceeded\n", nf);
#else
exit (BAD); /* apparently not */
#endif
}
if ((io.descr.d_stat & NETWRKD) == 0) /* networked ? */
{
printf ("Notesfile %s is not networked\n", nf);
exit (BAD);
}
r = q = title;
while (*r) /* look for last '-' */
{
if (*r == '-')
q = r; /* this is new last */
r++; /* on to the next */
}
suffix = q; /* save the pointer for possible later use */
/* like misformed articles with "- (nf)" */
length = q - title; /* non-nulls before '-' */
if (streq (q, NFSUFFIX) != 0 &&
streq (q, OLDSUFFIX) != 0)
/* so is not nf generated */
{
/*
* we come back up to here if the article had the notesfile
* suffix and was not properly formatted.
* This is a stopgap measure due to the fact that I misdesigned
* the system so that notes used a right associative rule
* to determine its articles, while news programs used a
* left associative rule. Oh well.
* - Ray Essick July 27, 1982
*/
badform: /* come here if no "#" in a "- (nf)" article */
getperms (&io, 1, origsys);
if (allow (&io, WRITOK) == 0)
{
printf ("System %s not allowed to write notes\n", origsys);
exit (BAD);
}
strmove (origsys, newsid.sys); /* build uniq id */
newsid.uniqid = (-newsseq * 100) - 1;
/* is not there yet */
strmove (origsys, note.n_id.sys);
note.n_id.uniqid = (-100 * newsseq); /* build the note descriptor */
copydate (&entered, ¬e.n_date);
strmove (fromsys, note.n_from);
stat = FRMNEWS; /* came from news system */
i = 0; /* fix up the title */
while (title[i++]); /* get to the end */
for (i--; i < TITLEN; i++)
title[i] = ' '; /* space pad */
lock (&io, 'n');
if ((notenum = chknote (&io, ¬e.n_id, ¬e2)) == 0)
{
if (!strncmp (title, "Re:", 3)) /* see if a "followup" */
{
p = title;
do
{
p + = 3;
while (*p == ' ')
p++; /* Skip Spaces */
} while (!strncmp (p, "Re:", 3)); /* get all re's */
strcpy (io.xstring, p);
notenum = findtitle (&io, io.descr.d_nnote);
if (notenum > 0 && !chkresp (&io, &newsid, ¬e2, notenum))
{
pagein (&io, rawnews, &where);
gettime (&whentime);
putresp (&io, &where, stat, notenum, &entered, &auth,
¬e, NOLOCKIT, &newsid, NOADDID, fromsys, ADDTIME,
&whentime);
}
else
if (notenum < 1)
{
pagein (&io, rawnews, &where);
notenum = putnote (&io, &where, title, stat, ¬e, &auth,
NOPOLICY, NOLOCKIT, NOADDID, fromsys, ADDTIME);
io.nnotrcvd++; /* count as networked in */
}
else
{
printf ("Duplicate Response handed back by news\n");
}
}
else
{
pagein (&io, rawnews, &where);
notenum = putnote (&io, &where, title, stat, ¬e, &auth,
NOPOLICY, NOLOCKIT, NOADDID, fromsys, ADDTIME);
io.nnotrcvd++; /* count as networked in */
}
}
else
if (note2.n_stat & ORPHND) /* replace foste parent */
{
pagein (&io, rawnews, ¬e2.n_addr); /* collect text */
gettime (¬e2.n_rcvd); /* current tod */
gettime (¬e2.n_lmod); /* last touched */
copyauth (&auth, ¬e2.n_auth); /* fill in the author */
note2.n_stat | = FRMNEWS; /* brand it */
for (i = 0; i < TITLEN; i++)
note2.ntitle[i] = title[i]; /* move new title */
copydate (&entered, ¬e2.n_date); /* written date */
strcpy (note2.n_from, fromsys); /* and who gave it to us */
putnrec (&io, notenum, ¬e2); /* and replace */
}
else
{
printf ("Duplicate news article received\n");
}
unlock (&io, 'n');
}
else
{ /* nf generated article */
sprintf (fname, "/tmp/nfn%d", getpid ()); /* generate name */
scr = fopen (fname, "w"); /* open scratch file */
while ((c = getc (rawnews)) != '#' && c != EOF) /* find start */
if (scr != NULL)
putc (c, scr); /* hold it in the scratch file */
if (c == EOF)
{
fclose (scr); /* flush what is there */
if ((rawnews = fopen (fname, "r")) == NULL)
{
printf ("Article lost in file system: %s\n", fname);
unlink (fname); /* remove the file */
closenf (&io); /* close the notesfile */
exit (BAD);
}
*--suffix = '\0'; /* remove " - (nf)" suffix */
unlink (fname); /* only link is the file descriptor */
goto badform; /* reparse as a news article */
}
fclose (scr); /* close and */
unlink (fname); /* toss the scratch file */
switch (getc (rawnews))
{
case 'N': /* base note coming through news */
if (fscanf (rawnews, ":%[^:]:%ld:%o:%d", note.n_id.sys,
¬e.n_id.uniqid, &status, &count) != 4)
goto toobad; /* bad form */
while (getc (rawnews) != '\n'); /* skip eol */
for (i = length; i < TITLEN; i++)
title[i] = ' '; /* space fill */
while ((c = getc (rawnews)) != '!'); /* skip sys name, got elsewhere */
for (i = 0; (i < NAMESZ - 1) && ((c = getc (rawnews)) != ' '); i++)
auth.aname[i] = c; /* get the author */
i = 0;
while ((line[i] = getc (rawnews)) != '\n');
/* grab date */
getdate (line, ¬e.n_date); /* use secret decoder ring */
while (getc (rawnews) != '\n'); /* loop in case trash on line */
/* in theory this line is empty */
getperms (&io, 1, note.n_id.sys); /* see if he's allowed */
if (allow (&io, WRITOK) == 0) /* not a chance */
{
closenf (&io);
exit (BAD);
}
lock (&io, 'n'); /* get exclusive mining rights */
if ((notenum = chknote (&io, ¬e.n_id, ¬e2)) == 0)
/* is it there? */
{
pagein (&io, rawnews, &where); /* grab text */
status | = FRMNEWS; /* came through news! */
strmove (fromsys, note.n_from);
putnote (&io, &where, title, status, ¬e, &auth, NOPOLICY,
NOLOCKIT, NOADDID, fromsys, ADDTIME);
io.nnotrcvd++; /* count it */
}
else
if ((note2.n_stat & ORPHND) && NOT (status & ORPHND))
{ /* extant is orphan, new isnt */
pagein (&io, rawnews, ¬e2.n_addr);
/* grab text */
gettime (¬e2.n_rcvd);
gettime (¬e2.n_lmod); /* time stamp it */
copyauth (&auth, ¬e2.n_auth);
/* put correct author */
note2.n_stat = status + FRMNEWS;
/* correct status */
for (i = 0; i < TITLEN; i++)
note2.ntitle[i] = title[i];
copydate (&entered, ¬e2.n_date);
strmove (fromsys, note2.n_from);
putnrec (&io, notenum, ¬e2); /* and replace */
}
else
{ /* duplicate */
printf ("Duplicate note handed back by news\n");
}
unlock (&io, 'n'); /* open to the rest of world */
break;
case 'R': /* response coming through news */
if (fscanf (rawnews, ":%[^:]:%ld:%[^:]:%ld:%o:%d", note.n_id.sys,
¬e.n_id.uniqid, respid.sys, &respid.uniqid, &status, &count) != 6)
goto toobad; /* bad form */
while (getc (rawnews) != '\n'); /* skip eol */
getperms (&io, 1, respid.sys); /* see if he's allowed */
if (allow (&io, RESPOK) == 0) /* not a chance */
{
closenf (&io);
exit (BAD);
}
while ((c = getc (rawnews)) != '!'); /* skip sys name, got elsewhere */
for (i = 0; (i < NAMESZ - 1) && ((c = getc (rawnews)) != ' '); i++)
auth.aname[i] = c; /* get the author */
i = 0;
while ((line[i] = getc (rawnews)) != '\n');
/* grab date */
getdate (line, &entered); /* use secret decoder ring */
while (getc (rawnews) != '\n'); /* loop in case trash on line */
/* in theory this line is empty */
lock (&io, 'n'); /* lock it up */
notenum = chknote (&io, ¬e.n_id, ¬e2);
if (notenum == 0) /* build foster parent */
{
printf ("Orphaned response handed in by news\n");
strmove (fromsys, note.n_from);
note.n_nresp = 0;
note.n_auth.aid = ANONUID;
strcpy (note.n_auth.aname, "Unknown");
copydate (&entered, ¬e.n_date);
gettime (&whentime); /* get current time */
status = ORPHND + FRMNEWS; /* combo there */
for (i = 0, p = "Orphaned Response"; (i < TITLEN) && *p; p++, i++)
note.ntitle[i] = *p;
for (; i < TITLEN; i++)
note.ntitle[i] = ' ';
/* pad */
where.addr = 0; /* no text */
notenum = putnote (&io, &where, note.ntitle, status,
¬e, ¬e.n_auth, NOPOLICY, NOLOCKIT, NOADDID, fromsys, ADDTIME);
io.norphans++; /* census of orphans */
getnrec (&io, notenum, ¬e2); /* get good one */
}
if (chkresp (&io, &respid, ¬e2, notenum) == 0)
{ /* none, insert it */
status | = FRMNEWS;
pagein (&io, rawnews, &where);
putresp (&io, &where, status, notenum, &entered, &auth,
¬e, NOLOCKIT, &respid, NOADDID, fromsys, ADDTIME, &whentime);
io.nrsprcvd++; /* count him in */
}
else
{
printf ("Duplicate response handed back by news\n");
}
unlock (&io, 'n');
break;
default: /* bad news coming through news */
toobad: /* label for bad format jumps */
printf ("Some sort of failure caused jump to toobad\n");
closenf (&io);
exit (BAD);
}
} /* end of nf coming back in */
lock (&io, 'n');
getdscr (&io, &io.descr);
io.descr.netwrkins++; /* count as net in */
putdscr (&io, &io.descr);
unlock (&io, 'n');
finish (&io);
exit (GOOD);
}essick (07/30/82)
#R:cmcl2:-808200:uiucdcs:23300012:004:2625
uiucdcs!essick Jul 30 09:51:00 1982
This is a modified version of "tseach.c" which enables the
newsinput routines to use it when looking for the parent of a
response generated in news. The changes merely separated the gathering
of the search string from the actual search.
Thanks go to Malcolm Slaney (pur-ee!malcolm) for the changes.
Ray Essick
- - - - - -
#include "structs.h"
/*
* tsearch(io, fromnum, grabstring)
* search notetitles from note #fromnum back towards 1
* looking for the string in io->xstring. If that string is
* empty or grabstring is true, prompt the user for a string.
*
* Returns: 0 if searched and not found
* >0 the note number which matched
* -1 null string to search for.
*
* Original coding: Ray Essick January 1982
* Modified: Malcolm Slaney July 1982
* to separate the gathering of a string and
* the actual search. A Good plan. (RBE)
*/
tsearch (io, fromnum, grabstring)
struct io_f *io;
{
register int i;
if (grabstring || io -> xstring[0] == '\0')
{
at (-1, 1);
printf ("Search String: ");
i = gline (io -> xstring, TITLEN); /* grab one */
at (-1, 1);
printf ("%*s", i + 16, " "); /* clear the line */
if (i == 1)
{
io -> xstring[0] = '\0';
return - 1;
}
}
at (0, 1);
fromnum = findtitle (io, fromnum);
if (intflag)
{
intflag = 0; /* don't recatch this one */
printf ("Search aborted");
}
if (fromnum > 0)
return (fromnum);
else
printf ("%s: Not Found", io -> xstring);
return 0;
}
findtitle (io, fromnum)
struct io_f *io;
{
struct note_f note; /* hold the note descriptor */
register int i,
j,
xlength;
for (j = 0; io -> xstring[j]; j++)
io -> xstring[j] = tolcase (io -> xstring[j]);
if (io -> xstring[0] == '\0') /* if still empty then */
return - 1; /* he doesn't really wanna search */
xlength = 0;
i = 0;
while (io -> xstring[i++])
xlength++;
if (fromnum > io -> descr.d_nnote)
fromnum = io -> descr.d_nnote;
intflag = 0; /* catch interupts */
while (fromnum > 0)
{
getnrec (io, fromnum, ¬e); /* grab descriptor */
if (note.n_stat & DELETED)
{
fromnum--;
continue; /* skip this one */
}
for (j = 0; j < TITLEN; j++)
note.ntitle[j] = tolcase (note.ntitle[j]);
for (j = 0; j < TITLEN + 1 - xlength; j++)
{
for (i = 0; i < xlength; i++)
if (io -> xstring[i] != note.ntitle[j + i])
{
i = -1;
break;
}
if (i != -1)
return fromnum;
}
fromnum--; /* move on to the next one */
if (intflag)
break; /* interupt */
}
return (0);
}