[prev in list] [next in list] [prev in thread] [next in thread] 

List:       bsdi-users
Subject:    elm2.5.3
From:       Digby Tarvin <digbyt () acm ! org>
Date:       2000-01-31 19:11:29
[Download RAW message or body]


Greetings all,

I recently compiled and installed elm2.5.3 on my BSD/OS 3.1 system
to correct the Y2K problem in the distributed version, and thought
I would share my experience so far and ask if anyone else has
tried it.

It seemed to configure and build without problem. I based
the configuration on the settings used by BSDI on the original
contrib version, although it did look like SGID mode wasn't necessary
given the mode of /var/mail:
	drwxrwxrwt  2 root  mail  512 Jan 31 02:00 /var/mail
(unless this isn't the way it was installed originally..)

The first hint of a problem was that frequently it left my mailbox
un-updated, claiming a permission problem writing to /tmp/snd.xxx
when I tried to exit.

The problem seems to be that elm uses a huge amount of
memory when running. My mailbox sits at around 16MB, which may
be a little large, but ps shows the process as:

  UID   PID  PPID CPU PRI NI   VSZ  RSS WCHAN  STAT  TT       TIME COMMAND
  100   428   422   0  18  0   384  208 pause  IWs   p0    0:01.53 -ksh (ksh)
->100  3261   428  13  10  0 29000 19240 wait   IW+   p0    0:38.48 /usr/local/b
  100  3305  3261   9  10  0   228  208 wait   IW+   p0    0:00.13 sh -c /usr/b
  100  3309  3305   0   2  0   420  564 select I+    p0    0:01.98 /usr/bin/vi 
  100   922   921   1   3  0   384   88 ttyin  IWs+  p1    0:00.93 -ksh (ksh)
  100   379    60   2  18  0   384  344 pause  Ss    01    0:04.16 -ksh (ksh)
  100  3313   379   6  29  0   136  252 -      R+    01    0:00.00 ps -l

The problem with /tmp/snd.xxx turned out to be that elm was trying to
fork in order to seuid/setgid back to real before doing a trial
create, and the fork was failing. Seems like a horribly inneficient
way of doing it to me, and I guess this sort of programming explains
the other problem I have notices, which is that it is now horribly
slow... (is there any reason why using access() on the file
if it already exists, and on the directory if it doesn't, would
not be safe???)

I think my problems were caused by running out of swap, not by
individual user restrictions, but I didn't see any system messages
logged. Anyone know if 'systat -swap' gives an accurate indication
of system memory availability? My user limits are setup as follows:

	time(cpu-seconds)    unlimited
	file(blocks)         unlimited
	coredump(blocks)     unlimited
	data(kbytes)         65536
	stack(kbytes)        2048
	lockedmem(kbytes)    10161
	memory(kbytes)       30480
	nofiles(descriptors) 128
	processes            64
But I assume most of these are per-process, so would not cause a
fork to fail..

On the plus side, at least I can now add attachments without having
to switch to netscape to send the message, and generating a mailbox
name for saved messages based on the alias set up for that user
if it exists is useful - better than generating the name from
the senders email address, especially for mail from people 
with compuserve style numeric names....

Anyone else with experience? Does the version bundled with 4.1
have sensible date code without the bloat??

I suspect it is approaching time to bite the bullet and learn a
new mail MUA, but I was hoping to postpone that at least till
I have caught up with the OS updates....

Regards,
DigbyT
-- 
Digby R. S. Tarvin                                              digbyt@acm.org
http://www.cthulhu.dircon.co.uk

[prev in list] [next in list] [prev in thread] [next in thread]