dwho@nmtsun.nmt.edu (David Olix) (07/30/89)
Before you all send responses, comments, or whatever...
I think I finally figured the problem out. Apparently mini-inews (the version
distributed with nntp 1.5) calls the post() routine in nntpd to post it's
stuff. However, post() (post.c) calls spawn() (spawn.c) with the line:
retcode = spawn(inews, "inews", "-h", CONT_POST, ERR_POST, errbuff);
This forks off another mini-inews process, which then calls post(), which
forks off yet another mini-inews process, etc....
Questions:
1) Is this really what supposed to happen?
2) Is the following order for posting stuff correct?
Pnews => mini-inews => nntpd => (real) inews.
^ ^
| Note the difference here |
If so, I think this should be made *VERY* clear in the documentation as
having a machine crash 4 times in one hour due to process table overflows
is a real pain in the... *MMPH*!
--David Olix (dwho@nmtvax.nmt.edu)bin@primate.wisc.edu (Brain in Neutral) (07/31/89)
The inews supplied with the NNTP distribution ("mini-inews") should only
be installed on client machines (i.e., machines that don't have the news
spool directory on them). All it does is take the article, open a connection
to nntpd on the server machine (which should not be, as it apparently was in
your case, the same as the client machine), and dump the article into the
connection.
It is documented in the README file in the nntp/inews directory that mini-inews
expects to send news to a remote machine.
Try reinstalling the "real" inews from the news distribution.
Paul DuBois
dubois@primate.wisc.edu