In <199910171121.NAA05988@jdolecek.per4mance.cz>
dolecek@ics.muni.cz wrote:

> What is the right think to do for sun3 and and it's local definition
> of Debugger() in non-DDB case ? Should the Debugger() function be
> there even for non-DDB case ? IMHO it's use is limited in that
> case, since MI code calls Debugger() only when DDB is configured
> in.  MD code (like zs driver) can surely take advantage of fact that
> the functionality is always available. But it's use is at least
> inconsistent (dev/si.c & dev/fd.c call Debugger() only when DDB,
> other code uncoditionally).

Debugger() in sun3/stub.c seems to be used with KGDB.
Comments in locore.s says trap #15 is used for KGDB breakpoints.
sparc/zs.c uses zskgdb() instead of Debugger() in KGDB case,
and it seems more reasonable for me.
(Well, I'm not familiar with the kernel debuggers...)

> The contents of sun3/stub.c should be maybe moved do sun3/conf.c
> or sun3/machdep.c IMHO, but I'm not sure which staff is normally
> put to either of those files.

I check sun3/stub.c cvslog by cvsweb.netbsd.org and notice
it have contained various contents in every version.
(ether_output, memcpy, swapgeneric, mountroot, Debugger...)
I think now it should be removed, too.
(should be in sun3/kgdb_machdep.c?)
---
Izumi Tsutsui
tsutsui@ceres.dti.ne.jp