From:	CSBVAX::MRGATE!montnaro@steinmetz@SMTP 24-NOV-1987 13:50
To:	EVERHART
Subj:	Bad interactions between TO and Kept GNU Emacs

Received:  from sprite.moose (sprite.ARPA)
	by steinmetz.steinmetz (1.2/1.1x Steinmetz)
	 id AA18949; Tue, 24 Nov 87 13:42:23 est
Received: by sprite.moose (3.2/1.1x Steinmetz)
	id AA07840; Tue, 24 Nov 87 13:41:21 EST
Date: Tue, 24 Nov 87 13:41:21 EST
From: Skip Montanaro <montnaro@steinmetz>
Posted-Date: Tue, 24 Nov 87 13:41:21 EST
Message-Id: <8711241841.AA07840@sprite.moose>
To: gurus, morrow@crd, davidsen@crd, barnum@crd, sajeski@crd
Cc: wildes@crd
Subject: Bad interactions between TO and Kept GNU Emacs


Doug Wildes pointed out a problem to me yesterday that can affect people on
VMS VAxes who run GNU Emacs in a subprocess as their default editor when
they try to send mail using TO. It may have been reported before, but I
hadn't heard of it.

If users tell TO that EMACS is their editor, and it is defined to execute
GNU emacs in a subprocess, TO will obligingly spawn a subprocess to edit the
draft file. The user will be able to suspend that emacs process and return
to the TO prompt, without having to save the draft file. This may result in
a mail message that has an empty message body, or an incomplete one,
depending on whether the user saved the draft file at all in the editor
session.

To get around this, I defined a GNU Emacs Lisp command that prompts the user
to save modified files when suspending. Substitute-key-definition can be
used to rebind suspend-emacs keys to execute save-buffers-suspend-emacs when
pressed.

    (defun save-buffers-suspend-emacs (&optional arg)
      "Offer to save each buffer, then suspend this Emacs fork.
    With prefix arg, silently save all file-visiting buffers, then suspend."
      (interactive "P")
      (save-some-buffers arg t)
      (suspend-emacs))

    (substitute-key-definition
      'suspend-emacs
      'save-buffers-suspend-emacs
      (current-global-map))

I recommend users who use both TO and kept GNU Emacs add this Lisp code to
their sys$login:.emacs file.

I'm not sure what signal EMACS is receiving from TO or from VMS when the TO
command exits, but it does appear to be getting killed by one of them. If
EMACS receives a hangup signal (SIGHUP) it will checkpoint all modified
files before exiting. Consideration might be given to checking for child
subprocesses before TO exits, and sending them a hangup signal, rather than
leaving them prey to whatever fate is currently befalling them. That way,
users could at least retrieve their checkpoint files.

Skip