[Libvirt-cim] [PATCH 1 of 2] [TEST] Added the indications.py to support Indications

[Deepti B Kalakeri]

Kaitlin Rupert wrote:
> Deepti B. Kalakeri wrote:
>> # HG changeset patch
>> # User Deepti B.Kalakeri <deeptik at linux.vnet.ibm.com>
>> # Date 1247143660 25200
>> # Node ID 73aedcd4026381023be4d25d741ce74113b8f8e6
>> # Parent 7b18b827fca6916fd87ff417f79a96c8a19b4913
>> [TEST] Added the indications.py to support Indications.
>>
>> Included the following important functions in indications.py
>> - sub_ind() --> Subscribe for the indications to be watched.
>> - handle_request() --> Filters the required indications.
>> - poll_for_ind() --> Wait for the required indications to be triggered.
>>
>
> Oops, I think I was unclear in my previous mail. I think the 
> indication tester should stay intact... The reason we decided to copy 
> the indication_tester from libcmpiutil to cimtest is so that in bugs 
> that get fixed in one get fixed in the other. That's why they need to 
> stay in sync.
>
> So, if you find a bug in filter_xml(), you want to be sure its fixed 
> in both places. The problem was that cimtest wasn't publicly available 
> at the time, and indications aren't trivial to test using standard CIM 
> related tools. So we wanted to offer users a way to test their 
> indications, which is why it exists in libcmpiutil.
>
> We could remove the indication_tester.py from libcmpiutil and leave 
> the only copy in cimtest, however, I think we should have two copies 
> of the same code in cimtest.
I guess you meant to say "I think we not should have two copies" ???
>
> So, I suggest leave indication_tester.py as is and place the functions 
> below in their own file. Does that seem reasonable? That way, we don't 
> duplicate code - if we fix a bug in one spot, we won't have to fix the 
> same bug somewhere else.
ok sure. done..
>
>> +def sub_ind(ip, virt, ind_names):
>> + dict = set_default(ip)
>> + sub_list = {}
>> + port = 5
>> +
>> + for ind, iname in ind_names.iteritems():
>> + ind_name = get_typed_class(virt, iname)
>> +
>> + sub_name = "Test%s" % ind_name
>> + port += 1
>> +
>> + sub = CIMIndicationSubscription(sub_name, ind_name,
>> + dict['default_ns'],
>> + dict['default_print_ind'],
>> + dict['default_sysname'],
>> + port)
>> + sub.subscribe(dict['default_url'], dict['default_auth'])
>> + logger.info("Watching for %s", iname)
>> + ind_names[ind] = ind_name
>> + sub_list[ind] = sub
>> +
>> + return sub_list, ind_names, dict
>> +
>> +def handle_request(sub, ind_name, dict, exp_ind_ct):
>> + #sfcb delivers indications to all registrations, even if the 
>> indication
>> + #isn't what the registration was subscribed to. So, for modified 
>> and + #deleted indications, we must loop through until the indication 
>> we are
>> + #looking for is triggered.
>> + for i in range(0, exp_ind_ct):
>> + sub.server.handle_request() + if len(sub.server.indications) < 1:
>> + logger.error("No valid indications received")
>> + return FAIL
>> +
>> + if str(sub.server.indications[0]) == ind_name:
>> + sub.unsubscribe(dict['default_auth'])
>> + logger.info("Cancelling subscription for %s", ind_name)
>> + return PASS
>> + else:
>> + sub.server.indications.remove(sub.server.indications[0])
>> +
>> + logger.error("Did not recieve indication %s", ind_name)
>> + return FAIL
>> +
>> +def poll_for_ind(pid, ind_name, timeout=20):
>> + status = FAIL
>> + for i in range(0, timeout):
>> + pw = waitpid(pid, WNOHANG)
>> +
>> + # If pid exits, waitpid returns [pid, return_code] + # If pid is 
>> still running, waitpid returns [0, 0]
>> + # Only return a success if waitpid returns the expected pid
>> + # and the return code is 0.
>> + if pw[0] == pid and pw[1] == 0:
>> + logger.info("Great, got '%s' indication successfully", ind_name)
>> + status = PASS
>> + break
>> + elif pw[1] == 0 and i < timeout:
>> + if i % 10 == 0:
>> + logger.info("In child, waiting for '%s' indication", ind_name)
>> + sleep(1)
>> + else:
>> + # Time is up and waitpid never returned the expected pid
>> + if pw[0] != pid:
>> + logger.error("Waited too long for '%s' indication", ind_name)
>> + kill(pid, SIGKILL)
>> + else:
>> + logger.error("Received Indication error: '%d'", pw[1])
>> +
>> + status = FAIL
>> + break
>> +
>> + return status
>
>

-- 
Thanks and Regards,
Deepti B. Kalakeri
IBM Linux Technology Center
deeptik at linux.vnet.ibm.com