Path: utzoo!attcan!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!uwm.edu!rpi!bu.edu!nntp-read!jbw
From: jbw@bucsf.bu.edu (Joseph Wells)
Newsgroups: comp.emacs
Subject: Re: vi triumphs over emacs! ("bug" in emacs)
Message-ID: <JBW.90Sep16211004@bucsf.bu.edu>
Date: 17 Sep 90 01:10:04 GMT
References: <9009111909.AA00518@next-8.gac.edu> <45656@cornell.UUCP>
Sender: news@bu.edu.bu.edu
Organization: Boston University Computer Science Department
Lines: 71
In-reply-to: murthy@algron.cs.cornell.edu's message of 11 Sep 90 20:08:37 GMT

>>>>> "Scott" == scott@NIC.GAC.EDU
Scott> What happens is that I log into the system with a suitably large vt100
Scott> window, something like 80x72, and then eventually get into emacs to
Scott> use gnus.  Fine.  But, since almost without fail I forget to do my
Scott> stty rows call first, I end up with only a small window out of this
Scott> huge one.

Scott> So, what I attempt it ctrl-z out of emacs, stty rows 72, fg back in.
Scott> This doesn't fix it.  What I end up having to do is find the pid for
Scott> emacs, and do:

Scott> (sleep 5 ; kill -WINCH <emacs-pid>) &
Scott> fg

Scott> after the stty, to get emacs to recognize the window size change.

The following piece of GNU Emacs Lisp should handle your problem:

(setq emacs-pid
      (save-excursion
	(set-buffer (get-buffer-create " *Emacs PID*"))
	(erase-buffer)
	(call-process "/bin/sh" nil t nil "-c"
		      "ps -l$$")
	(goto-char (point-min))
	(let (case-fold-search)
	  (re-search-forward "[ \t]+PID")
	  (let ((col (current-column)))
	    (forward-line 1)
	    (move-to-column col))
	  (read (current-buffer)))))
(setq suspend-resume-hook
      (function
       (lambda ()
	 (call-process "kill" nil nil nil "-lWINCH" (format "%s" emacs-pid)))))

Scott> Meanwhile, when doing various stuff at the command line, I tend to use
Scott> vi (quicker, well, sort of).  In vi, I can ctrl-z out of the editting
Scott> session, stty rows 72, and fg back in, without a problem.  vi detects
Scott> the resize, and applies it.  Wonderful, none of this sleep/kill
Scott> hoopla.

GNU Emacs unfortunately only interrogates the system for the screen size
when it receives a SIGWINCH and on startup.

>>>>> "Chet" == murthy@algron.cs.cornell.edu (Chet Murthy)
Chet> As I understood it, the _CORRECT_ protocol is for the
Chet> terminal-emulator to signal the process group leader for the
Chet> slave-side of the pseudo-terminal with WINCH, to let it know that the
Chet> window size has changed.  That is what SIGWINCH was invented for.

That's exactly what's happening.  His shell is the process group leader of
the terminal's current process group.

Chet> So the problem seems to me to be with NeXT's terminal emulator, not
Chet> Emacs.

Nope.

Chet> But you're right - this is an easy problem to handle.  Of course, if
Chet> there were a way to get the process-id of Emacs easily (within Emacs),
Chet> and do a UNIX kill(2), you could just do the kill from within Emacs.
Chet> But apparently that has been left out.  About 2 seconds' hacking.

There is a way to get Emacs's pid without starting another process, if
CLASH_DETECTION was specified when Emacs was built.  Then you can lock a
file and look at the contents of the lock file to get the pid.  (Gross,
huh?)

-- 
Joe Wells <jbw@bu.edu>