Re: [Qemu-devel] [PATCH v3] trace: [qmp] Add QAPI/QMP commands to query and control event tracing state

[Lluís Vilanova]

Markus Armbruster writes:

> Lluís Vilanova <address@hidden> writes:
>> Eric Blake writes:
>> 
>>> On 08/21/2014 11:52 AM, Lluís Vilanova wrote:
[...]
>>>> +#
>>>> +# Since 2.2
>>>> +##
>>>> +{ 'command': 'trace-event-set-state',
>>>> +  'data': {'name': 'str', 'state': 'bool', '*keepgoing': 'bool'} }
>> 
>>> This only lets me set the state of one name at a time.  Oh, unless I'm
>>> setting a pattern, and it then sets the state of all names that match
>>> that pattern.  I'm wondering if you should have 'name':['str'] to allow
>>> me to set multiple names/patterns in one go, instead of having to call
>>> the command multiple times; but it's probably not worth the complexity.
>> 
>> I agree with the complexity comment. Also, the keepgoing is useful to set 
>> events
>> using a pattern, even if some of them are statically disabled (otherwise it
>> gives an error).

> Yes, let's keep this simple.

> However, I feel keepgoing is unnecessarily vague.  Its purpose is to
> enable use of a pattern in the presence of disabled events.  I'd prefer
> to nail it down to exactly that purpose rather than having it cover
> arbitrary, unspecified errors.

> A few ideas on how to do that:

> * Have a flag to modify the semantics of the pattern: either "match all
>   events", or "match just disabled and enabled events, not unavailable
>   events".

> * To find out what a trace-event-set-state actually does, you need to
>   trace-event-get-state the same pattern.  We could make
>   trace-event-set-state return the events it changed, so you never have
>   to trace-event-get-state, but it's probably not worthwhile.

> [...]

I've renamed it to "ignore-unavailable" (off by default).


Thanks,
  Lluis

-- 
 "And it's much the same thing with knowledge, for whenever you learn
 something new, the whole world becomes that much richer."
 -- The Princess of Pure Reason, as told by Norton Juster in The Phantom
 Tollbooth