🚀 go-pugleaf

The following event occurred in 18.57 emacs running on an IBM RS 6000
running AIX...

> Date: Tue, 27 Aug 91 15:05:33 -0400
> From: young
> To: Bug-EMACS, Bug-RS6000
> Subject: SIGDANGER kills subprocesses, 18.57.2
> 
> When AIX runs out of swap space, it sends a SIGDANGER signal ("you are

Actually, when you are running *low* on swap space.

> running low on swap space") to "suspect processes".  (Probably the big
> ones.)

Probably all processes get the signal (one hopes).  If they don't free
up space as a result of the SIGDANGER, the paging space monitor KILLS
the biggest processes first.  Naturally.

> 
> If EMACS receives one of these signals, it kills all its subprocesses.
> This is a big lose if one of those processes is The ICAD System.

(The ICAD System <=> a huge lisp program - Franz Allegro Common Lisp)

> 
> Note that this only happens on the /first/ SIGDANGER signal.  The
> symptom in a subshell is, e.g.:
> 
> 	Process cmushell hangup
> 
> 				--- Jonathan

I tried this personally, doing a KILL -DANGER %emacs and emacs bleated
from the background:

Fatal error (33).

It is true that successive SIGDANGER's don't do any apparent harm, so
it seems likely to me that this is deliberate behaviour on EMACS'
part, and someone just forgets to reset the signal handler.  I
guess the default behaviour for SIGDANGER must be to ignore it. 

I cannot investigate this any further at the moment, but I hope you
can suggest some sort of fix- it would be best if emacs issued a
warning to the user and then did nothing, since the page space monitor
will be sending the SIGDANGER to the child processes too.  Killing off
the children without warning is definitely *not* the right thing to
do.


 thanks,

  --nick caruso
    icad inc, 201 broadway			   
    peoples republic of cambridge, MA 02139-1901	   
    internet: nick@icad.com / UUCP: ...uunet!icad.com!nick	   
    ------------------------------------------------------