[libvirt] [PATCH 1/9] Introduce node device lifecycle event APIs

[Cole Robinson]

On 07/26/2016 01:32 PM, Daniel P. Berrange wrote:
> On Tue, Jul 26, 2016 at 01:28:34PM -0400, Cole Robinson wrote:
>> On 07/20/2016 09:50 AM, Jovanka Gulicoska wrote:
>>> Node device lifecycle event API entry points for registering and deregistering
>>> node deivce events, as well as types of events associated with node devices.
>>
>> * device
>>
>>> These entry points will be used for implementing asynchronous lifecycle events.
>>
>> These three lines are a bit long. Maybe split them to be 70-75 chars:
>>
>> http://stackoverflow.com/questions/2290016/git-commit-messages-50-72-formatting
>>
>>>
>>> Node device API:
>>> virConnectNodeDeviceEventRegisterAny
>>> virConnectNodeDeviceEventDeregisterAny
>>> virNodeDeviceEventLifecycleType which has events CREATED and DELETED
>>
>> So one major question here is that we are only added CREATED and DELETED
>> events, but devices can also be 'updated', as in their config can change. For
>> example when a cdrom device has media ejected or inserted, udev fires an
>> event, and we update the cached device config which is reflected in the device
>> XML.
>>
>> Adding an UPDATED event or similar isn't hard, but it's not technically a
>> lifecycle. So if we add it, it should be a separate callback? Similar to what
>> was eventually done for the pool refresh event. CCing danpb for his thoughts
> 
> Yep, I'd recommend a separate event callback for that.
> 

Sounds good.

Jovanka, I suggest just sticking a patch at the end of the series which adds
all the UPDATED handling in one shot, rather than break it into a bunch of
patches. Similar to what Dan did for the REFRESHED event:

http://www.redhat.com/archives/libvir-list/2016-June/msg01897.html

Note there's not a clear way to unit test this, since the test driver doesn't
really have any plumbing for it, so those bits can be skipped.

Thanks,
Cole