On Thu, 17 May 2012, Kurt Geebelen wrote:

>
>
> Op 17/05/2012 12:15, Herman Bruyninckx schreef:
>> On Wed, 16 May 2012, Kurt Geebelen wrote:
>>
>>>
>>>
>>> Op 16/05/2012 18:28, Ruben Smits schreef:
>>>> On Wed, May 16, 2012 at 5:30 PM, Kurt Geebelen
>>>> <kurt [dot] geebelen [..] ...> wrote:
>>>>>
>>>>> Op 16/05/2012 17:24, Charles Lesire-Cabaniols schreef:
>>>>>
>>>>>
>>>>>
>>>>> 2012/5/16 Peter Soetens<peter [..] ...>
>>>>>> On Wed, May 16, 2012 at 2:51 PM, Kurt Geebelen
>>>>>> <kurt [dot] geebelen [..] ...> wrote:
>>>>>>> Hi all,
>>>>>>>
>>>>>>> I have 2 components A and B. They are connected via a port P of type
>>>>>>> std::vector<double>. It is an outputport in component A, and an
>>>>>>> inputport in component B.
>>>>>>>
>>>>>>> In component A, I write to the port in startHook, and then write to it
>>>>>>> sporadically.
>>>>>>>
>>>>>>> Component B is being triggered by an other port, but when it is
>>>>>>> triggered, it should read port P. It doesn't matter if the data on
>>>>>>> port
>>>>>>> P is old or new.
>>>>>>>
>>>>>>> Now what is happening that the read-command returns NoData, where I
>>>>>>> would expect it to return OldData. I checked in the deployer the value
>>>>>>> of port P in component A, and there really is a value on the port.
>>>>>>>
>>>>>>> Any idea why NoData is returned?
>>>>>
>>>>> Maybe you could send us at least your deployment script?
>>>>>
>>>>> Deployment script is on attachment. The connection that is giving
>>>>> problems
>>>>> is on line 101:
>>>>> connect("dynamicMHE.portMeasurementsCtrl","mcuHandler.controlOutputPort",cp)
>>>>> (mcuHandler is the component which outputs, dynamicMHE takes the input.
>>>>> None
>>>>> of the other components use mcuHandler.controlOutputPort)
>>>>>
>>>>> But I found a way in which I don't have the problem: if I start
>>>>> mcuHandler
>>>>> at the end of the script, e.g. on line 91, in stead of in the beginning
>>>>> (now
>>>>> on line 23) I do get OldData from the port, and not NoData.
>>>>>
>>>>> This thus solves my problem, but I find it weird why this happens..
>>>>>
>>>> It seems completely normal to me, since you create the connection
>>>> after the write (which happens in your startHook). The data never
>>>> arrived in the receiving component.
>>> Ah, ok, I see.
>>>
>>> Thanks,
>>> Kurt
>>
>> A piece of advice: separate deployment from runtime use. I mean: use the
>> deployment script to get all your components up, but not yet running. Once
>> they are all ready, only then you let your coordinator send out a run event
>> after which data can begin to flow.
>>
>> This is the logic of having (at least) four separate phases in the
>> "development process":
>> - component development
>> - system building
>> - deployment
>> - runtime
>> Threat them differently, and don't use files or scripts that mix two of
>> them.
>
> That's exactly what we did yesterday after discovering the behavior :-) .

Good! These kind of things are the essence of what we try to make more
clear and document in the BRICS project ("Best Practice in RobotICS"). So,
your experience is a corroboration for me that this piece of advise is
indeed best practice :-)

Herman