Path: utzoo!attcan!uunet!cs.utexas.edu!rutgers!cbmvax!vu-vlsi!dsinc!syd
From: syd@DSI.COM (Syd Weinstein)
Newsgroups: comp.mail.elm
Subject: Re: Elm 2.2 PL10
Keywords: All mail is lost if /usr/spool is full !
Message-ID: <1989Nov16.145450.14312@DSI.COM>
Date: 16 Nov 89 14:54:50 GMT
References: <91266@pyramid.pyramid.com>
Reply-To: syd@DSI.COM
Organization: Datacomp Systems, Inc.  Huntingdon Valley, PA
Lines: 67

alh@pyrnova (Alan Holzman) writes:

:In article <90691@pyramid.pyramid.com> alh@pyrnova.pyramid.com (Alan Holzman) writes:
:>
:>	I've just seen a second occurence of the following :
:>
:>	I finish reading/reviewing my mail and am ready to exit. I often
:>	leave between 3-12 active messages in my mailbox when I quit, rather
:>	than having them saved to the =received folder. When I did a "q" to
:>	quit Elm, a saw a system message that /usr/spool was full. When I
:>	tried to then re-enter Elm or even use bsd Mail, all of those 
:>	messages were gone ! 
:Syd ;
:	Could you or someone else please respond to this and the number of
:other postings regarding this problem. If there is currently no fix available,
:or if a fix is in the works for x weeks / months in the future and also what
:your recommendation is for today. Should we all do as I suggest in item # 4
:or are their other workarounds ?

Ok, why I don't respond each time this crops up is that it is in
the monthly posting:

EB09 Elm(1) does not check for failed writes and closes of files.
     This can have a disastrous affect if /tmp runs out of space.
     In this case, elm(1) might copy a mailbox to a /tmp file and
     then back.  Since elm(1) wouldn't detect that the copy to
     the temp file failed, the temp file would wind up empty,
     thereby causing the copy back to the mailbox to lose all
     messages.


This is an 'OLD' bug, as you can tell by its low number.  There is
no decent workaround.  In fact, without a complete rewrite to Elm,
trying to fix it is a patchwork problem, you might get some of the
places it occurs, but not all of them.

I have received one patch that has yet to be processed that addresses
part of the problem.  I have yet to reveiw it to see how much of the
problem it addresses.

Elm can run into trouble and loose all your messages if either
/tmp runs out of space or /usr/spool/mail (or /usr/mail) runs
out of space.  If /tmp runs out of space, an emergency exit ('x' command)
will at least not loose the messages, as it won't do a copy back.
If /usr/spool/mail runs out of space, the problem is worse, as mail
is going to get lost anyway, as there is no room to append to your mailbox
for incoming messages either.

The best workaround is to try and manage things so that those file systems
don't run out of space, and I know that can be difficult.  I am open
to patches (please let me get the one I have out first so you don't
duplicate effort) that solve the problem.  However, no matter how major
you feel it is, its a lower priority one on my books than the misaddressing
problem:

EB26 When using an address of the form "node!user@domain", the !
     has higher precedence than the @ when elm resolves the ad-
     dress.  The @ should have higher precedence.

I feel this way since proper administration of the file system sizes
should prevent running out of space in /usr/spool/mail.  However,
EB26 causes mail to go astray.
-- 
=====================================================================
Sydney S. Weinstein, CDP, CCP                   Elm Coordinator
Datacomp Systems, Inc.				Voice: (215) 947-9900
syd@DSI.COM or {bpa,vu-vlsi}!dsinc!syd	        FAX:   (215) 938-0235