Steve Peng writes:
>
> Tom,
>
> Thanks for your comments. Responses are inline.
>
> Steve
>
>
> Tom Whitten wrote:
> > Steve Peng writes:
[SNIP]
> >
> > usr/src/lib/libscf/inc/libscf_priv.h:
> > - I think the comment at line 295 "Switch client" gives the wrong
> > impresion. I think that it is the repository that is being
> > switched -- not the client.
> >
> >
> So it is a client which does repository switch. How about "Repository
> switch client"?
Ok.
[SNIP]
>
> > usr/src/cmd/svc/configd/backend.c:
> > - lines 1102 to 1205
> > I encourage you to not use the REP_PROTOCOL_FAIL_UNKNOWN return
> > value. That does not provide much informatin to the client. If
> > none of the existing error codes are appropriate, perhaps we
> > should create a new one.
> >
> Ah! I thought that one too when I implemented changes and decided to
> use what had been used in the backup client.
> I am guessing the reason we use that in the current repository backup
> client is that the error message will be sent to the console from the
> underlying filesystem not from the smf framework if the operation
> fails. Of course we can print out additional smf specific error message
> to the console but I think the message from filesystem may be sufficient
> enough. I am ok to create
> a new error message but that will mean we may need to obsolete the one
> used in the backup client as well to be in sync.
I think that you are correct. You also have the calls to
configd_critical() to log information about the error.
>
[SNIP]
> > usr/src/common/svc/repcache_protocol.h:
> > - line 356
> > Need to document the new request type.
> >
> Which line were you referring to? I did not see anything but '*'.
Exactly. All of the repcache protocol messages are documented in this
comment except for your new message. In my obscure way I was trying to say
that you should add the SWITCH message to the block comment.
>
[SNIP]
