Thu, Jan 03, 2002 at 08:51:43AM -0600, Ron Roskens:
> 
> A new method for catching a "break" signal was added in -current.
> 
> man cnmagic

does this actually work, for anyone?  eg:

i386% sysctl hw.cnmagic | od -a
0000000    h   w   .   c   n   m   a   g   i   c  sp   =  sp   ' soh  nl

which i think is ' ctrl-A; but that does not dump it into ddb.  even change
the sequence:

i386% sysctl -w hw.cnmagic='+++'
hw.cnmagic: '` -> +++
i386% sysctl hw.cnmagic | od -a
0000000    h   w   .   c   n   m   a   g   i   c  sp   =  sp   +   +   +

still no ddb.  what's interesting is that the console does not echo +'s
after the second.  it doesnt work for recent sparc or sparc64 either.


> On Thu, 3 Jan 2002, Simon J. Gerraty wrote:
> 
> > Date: Thu, 3 Jan 2002 02:52:16 -0800 (PST)
> > From: Simon J. Gerraty <sjg@crufty.net>
> > To: port-sparc@netbsd.org
> > Cc: sjg@crufty.net
> > Subject: BREAK on console ignored
> >
> > I'm not exactly sure when this stopped working...
> >
> > # uname -a
> > NetBSD too.crufty.net 1.5ZA NetBSD 1.5ZA (TOO) #0: Tue Jan  1 03:00:52 PST 2002     root@too.crufty.net:/l0/sys/compile/TOO sparc
> >
> > This is an SS20 with serial console.
> >
> > During initialization, BREAK drops me to the prom's ok prompt.
> >
> > Once the kernel has started booting though, BREAK is ignored.
> > My normal kernel does not have either of DDB or KGDB defined.
> > From dmesg.boot:
> >
> > zstty0 at zs0 channel 0 (console i/o)
> >
> > suggests that zs is being used for the console.  Here I see zs_abort()
> > which looks like it should end up calling prom_abort, but I don't see
> > where zs_abort is hooked in.
> >
> > If I'm off track, a pointer would be appreciated.
> >
> > Thanks
> > --sjg
> >
> >
> 
> Elfin