🚀 go-pugleaf

One of the users here reports that emacs seems to be losing track of
processes.  Here's his initial bug report:

		      *************************

From: acevedo@ATHENA.MIT.EDU
Date: Thu, 21 Mar 91 23:41:29 EST
To: jik@ATHENA.MIT.EDU
Subject: emacsdev


I have a weird problem with the emacs in emacsdev losing track of a
subprocess. I noticed it with the znol.el thingie I wrote. 

I wrote a decent znol.el ... lets you control users put in there (i.e.,
you can hook it to an emacs zwgc mode so when someone logs in/out it
shows up in the *znol* buffer. You can also divide the people into
groups, i.e., "Friends", "Consultants", "AVOID!"  :-)  etc. 

Anyway, part of the functionality is that it does like xznol; it
rechecks every so often. To implement this, I run znol; then, once the
process exits, I run a sleep for n seconds. After that exits, I re-run
znol, plus another sleep, continuing the loop. 

However, when I'm running my full emacs -- with zwgc, eolcr,
display-time -- and I do this, emacs tends to lose track of the sleep
process. It just gets started, and exits somehow, but the sentinel
doesn't get called when it does. (I now it exited at least somewhat
cleanly because I did a (setq foo (start-process ...)), and check the
process-status of foo ... it was 'exit.) Unfortunately, I can't reliably
reproduce it short of loading my emacs, and loading znol.el. You can try
it ... it's in ~acevedo/Emacs/znol.el. Do (setq znol-retry-interval 5)
(that's in seconds) so you don't have to wait very long. I also noticed
that while running zwgc and display time, I would have this problem for
intervals of under one minute; above it was fine. Then I loaded eolcr,
with the zephyr client in the background, and it no longer worked. 

I'm really tired, having spent a couple hours trying to figure this out.
I'll try again tomorrow, and if I have any more info, I'll let you know.


	Raul

P.S. I tried this with the emacs on the packs, and it seems fine ...
	again, I'll double check on this tomorrow. 

		      *************************

I asked him to provide sample elisp code that illustrated the problem.
Here's his response:

		      *************************

From: acevedo@ATHENA.MIT.EDU
Date: Wed, 24 Apr 91 14:44:29 EDT
To: jik@pit-manager.MIT.EDU
In-Reply-To: "Jonathan I. Kamens"'s message of Wed, 24 Apr 91 09:47:46 -0400 <9104241347.AA15371@pit-manager.MIT.EDU>
Subject: Re: emacsdev emacs losing track of processes


  Have you managed to figure out any way to duplicate the problem that i
  could submit?  (I.e. znol.el doesn't count :-)

Ok, here's something:

  (defun test ()
    (interactive)
    (set-process-sentinel
     (start-process "sleep" nil "sleep" "2")
     'sleep-sentinel)
    (message "Yawn!  Napping for 2 seconds ..."))

  (defun sleep-sentinel (process status)
    (beep t)
    (message "Finished!"))

Unfortunately, it only reproduces it sometimes, not all the time.  This
is a really random error.  One thing I think I've noticed is that, if
you do M-x test, and it works properly, DO NOT try it again right away.
Switch buffers, open a new window and then kill it, do something, then
try it again.  (I'm not sure if you need to do some display thing ---
like switch buffers or create a new window --- for it to happen, or if
you can just move the cursor around a bit.)  Also, when it does happen,
it is mostly a problem that emacs won't know it has finished unless you
hit a key, or do something explicit like that; but sometimes it just
takes about 5 to 10 seconds, instead of forever or until you hit a key
(or even if the mouse leaves the screen).

	Raul

		      *************************

Jonathan Kamens			              USnail:
MIT Project Athena				11 Ashford Terrace
jik@Athena.MIT.EDU				Allston, MA  02134
Office: 617-253-8085			      Home: 617-782-0710