BIND 10 #2676: Provide ccsession::rpc_call and replace usual calls with it

[BIND 10 Development]

#2676: Provide ccsession::rpc_call and replace usual calls with it
-------------------------------------+-------------------------------------
            Reporter:  vorner        |                        Owner:
                Type:  enhancement   |  jinmei
            Priority:  medium        |                       Status:
           Component:  Inter-module  |  reviewing
  communication                      |                    Milestone:
            Keywords:                |  Sprint-20130305
           Sensitive:  0             |                   Resolution:
         Sub-Project:  Core          |                 CVSS Scoring:
Estimated Difficulty:  2.5           |              Defect Severity:  N/A
         Total Hours:  0             |  Feature Depending on Ticket:
                                     |          Add Hours to Ticket:  0
                                     |                    Internal?:  0
-------------------------------------+-------------------------------------
Changes (by vorner):

 * owner:  vorner => jinmei


Comment:

 Hello

 Replying to [comment:10 jinmei]:
 > It depends on the amount of work, which I'm not really sure about.
 > But my personal impression at that moment is that it's better to
 > complete this task and create another one for it.

 I'll create a ticket. I think it should not be more work than I've spend
 hunting the problem down, but it might turn out it doesn't work.

 > At least ddns confirmed auth was running at the time of its startup,
 > so it's reasonable that it expects to get a response from auth.  DDNS
 > also confirms the updated zone is primary (in an ad hoc way though,
 > and it's still true that xfrout may not be running).  Also, although
 > not the point of this discussion, a "hidden master" can't work without
 > auth anyway; xfrout doesn't accept xfr requests directly, at least
 > right now.

 Talking about that, I believe that the fact I need to run auth to be able
 to do DDNS or xfrout is wrong too. Also, the current architecture doesn't
 really allow multiple DDNS or xfrout instances. I'll add a note to the
 port-sharing research ticket, since I think it is tightly related.

 > I guess the revised code highlights the concern more clearly.
 > Consider what happens in `type(value)` after rpc_call raises
 > `RPCError`.

 Ah, yes, crap. I changed that so it would work, I hope, as before. I'd add
 a test confirming that, but I must admit I don't understand the code
 enough to be sure about what should be tested :-|.

 > I'm okay with keeping the code without converting the exception per
 > se.  This should have caught tests (and made us think about it),
 > though; it suggests stats are only poorly tested (not an issue of this
 > branch itself).

 I think this can be said about a lot of our python code. The libraries are
 little bit better tested, but the top-level module scripts (the things in
 bin/) are quite untested. I guess the only exception there is the ddns.
 And not only is the code not tested, it seems to be broken in many places.

 > - apparently I forgot to note this before, but see (and confirm)
 >   3bec0f1.  We should now be able to remove that comment.

 Yes, it seems so.

-- 
Ticket URL: <http://bind10.isc.org/ticket/2676#comment:12>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development