From owner-cvs-sys  Mon Mar  4 06:27:32 1996
Return-Path: owner-cvs-sys
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.3/8.7.3) id GAA02993
          for cvs-sys-outgoing; Mon, 4 Mar 1996 06:27:32 -0800 (PST)
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19])
          by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id GAA02988
          Mon, 4 Mar 1996 06:27:27 -0800 (PST)
Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.12/8.6.9) id BAA16598; Tue, 5 Mar 1996 01:22:32 +1100
Date: Tue, 5 Mar 1996 01:22:32 +1100
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199603041422.BAA16598@godzilla.zeta.org.au>
To: bde@zeta.org.au, davidg@Root.COM
Subject: Re: cvs commit: src/sys/i386/conf LINT src/sys/kern subr_prf.c
Cc: CVS-committers@freefall.freebsd.org, cvs-all@freefall.freebsd.org,
        cvs-sys@freefall.freebsd.org, gpalmer@freefall.freebsd.org,
        j@uriah.heep.sax.de, peter@jhome.dialix.com
Sender: owner-cvs-sys@FreeBSD.ORG
Precedence: bulk

>>Using a single variable to control debugger entries is actually not
>>simple.  Debugger() is called from 18 modules and you don't want to
>>have to change all of these and all future callers to test the variable
>>and decide what to do.  The variable should probably be tested in
>>Debugger() (panic if it is set), but then you need a separate entry
>>point(s) to force ordinary entry to ddb from the console(s).
>>Similarly for other entries to ddb (for stray trace traps, breakpoints,
>>and other exceptions).  Similarly if cons_unavail is set.

>   All that we really need is a "ddb_on_panic" sysctl variable. The other
>cases I (and I'm sure most others) don't care about.

panic() usually isn't called until after ddb returns (if ddb is installed).
E.g., trap() calls kdb_trap() in ddb for all fatal traps.  Thus preventing
panic() from calling Debugger() doesn't help much.

Bruce