Subject: Re: /kern/kernel
To: None <perry@piermont.com>
From: Robert Elz <kre@munnari.OZ.AU>
List: current-users
Date: 09/14/1998 09:46:26
    Date:        Sun, 13 Sep 1998 18:14:26 -0400
    From:        "Perry E. Metzger" <perry@piermont.com>
    Message-ID:  <199809132214.SAA01635@jekyll.piermont.com>

    Fortunately or unfortunately, at the rate NetBSD kernel development
    proceeds, lots of systems hackers are major users of these tools,

I've done my share of crash dump anaylsys over the years, I suspect
I used "ps" about 3 or 4 times total for that (on a crash dump), and
at least a couple of those were of the nature "look what i can do"
rather than anything productive.   I'm pretty sure I have never used
"w" or "top" or anything like that on a crash dump.   I guess vmstat
might occasionally be useful, but a gdb script (it was adb in my day,
same effect I presume) is likely to produce the useful numbers just as
well, if perhaps not in quite such a pretty format.

In any case, any tools that are going to be useful to kernel hackers
will be maintained in a state that they work, the kernel people will make
sure of that - and no-one is suggesting that should not be done.  It is
just that attempting to invent a new (object filename independant) way of
getting offsets into /dev/kmem just has to be an unproductive way to
spend time.   If there is to be a change at all, let it be a significant
step forward, not just another wart.

kre