
		43) pkactive never gets decremented


From: harpo!eagle!allegra!linus!smk
Date: Wed May 18 15:33:32 1983
Subject: guess what?  A possible bug in uucp (oh no, again!)
Newsgroups: net.bugs.uucp

In the uucp we got from 4.1bsd, the pkopen in pk.c has a line like:

	if (++pkactive >= NPLINES)

(and pkactive is defined in <sys/pk.h>.  This is the only time
pkactive is used (it's a global int in pk.h, so it's assumed to
be initialized to 0) in the uucp code.  IT NEVER GETS DECREMENTED!
Should there be a decrement in pkclose() in pk0.c?  What does
this bug mean?  Could it be the cause of uucp every once in awhile
failing?

	Thanks for this bug detection go to allegra!honey.


From: tektronix!zehntel!ihnp4!houxm!mhuxa!mhuxj!mhuxt!eagle!allegra!honey
Date: Thu May 19 10:02:35 1983
Subject: pkactive bug
Newsgroups: net.bugs.uucp


the pkactive bug gets you when there is work pending for lots of
sites.  when the daemon kicks up, each successful pkopen increments
pkactive.  peeking at pkopen(), it's clear that after NPLINES of them,
future pkopen's will fail.  this used to happen to allegra when she
started up the late night poll; the symptom is a bunch of PKSTART
FAILEDs in the LOGFILE.  i upped NPLINES to 20 or so as a patch, which
squanders space, but that's what vm is all about, right?  i informed
chesson, who probably has other things on his mind out at btl-west.
	peter honeyman

