Re: [certi-dev] Where are region extents handled?

[Mattias Helsing]

Hi Eric, all,
Hmmm, yes I see now that I've been fiddling with generated code. Sorry
about that...

I can see where the generation fails (in GenMsgCXX.py line 666 and
more) to generate proper serialization for M_DDM_Modify/Delete_Region
messages, but haven't tried to understand the python generator or
attempted any fixes.

The other thing I found was that the libRTI/hla-1_3/RTIambassador.cc
(in notifyAboutRegionModification(...)) sets the extents in the
request using the api in BasicMessage
(setExtents(std::vector<Extent>&), while my (naive) fix tried to
implement serialization using the members and api in the
M_Ddm_Modify_Region class, which is never set. It also appears that
the hla1_3 RTIambassador is the only code which uses the BasicMessage
api to set/get Extents, but I'm not sure about this.

Ok, I will now wait for some advice or help on how this should be
fixed. My guess is that
   a) the python code needs some help to understand how to generate
(de)serialization code for std::vector<Extent>
   b) RTIAmbassador should use M_Ddm_Modify_Region::setExtentSet(..)
over BasicMessage::setExtent(..)
   c) Possibly remove region and extents members and methods from
BasicMessage (???)

cheers
Mattias

On Thu, May 14, 2015 at 12:25 PM, Eric Noulard <address@hidden> wrote:
>
>
> 2015-05-13 23:25 GMT+02:00 Mattias Helsing <address@hidden>:
>>
>> Hi Eric, all
>>
>> We are trying to get regions to work with CERTI RTI (1.3) and are a
>> bit perplaxed on how to proceed.
>
>
> Hi Mattias,
>
> DDM and CERTI is ancien story and I should say that I never played with it.
> The latest use of DDM with CERTI I am aware of was around 2005.
> This is 2 year before I even started to work on CERTI.
>
> Moreover the DDM implementation of CERTI was done at a time when
> DDM spec themselves in the HLA 1.3 was a moving target.
> Pierre may give you more precise informations about DDM history in CERTI
> and in HLA in general.
>
> Since 2007, lots of code rework has been done, e.g. CMake use for
> cross-platform build,
> use an internal (partially generated) messaging system (BasicMessage
> hierarchy) in order to bring
> heterogeneous simulation support, multi-standard API support,
> Java and Python bindings  etc...
>
> Since 2007 (when I started working on CERTI) we did not make
> some effort to ensure that DDM was still working.
> I know it may seem a pitty but our work on CERTI was driven by the need
> of the projects using it. There has been no interest in DDM since 2007.
>
> So basically you may discover that DDM is not working at all in the current
> code base.
>
>
>>
>> Our sim have configurable extent(s)
>> and have initially tried to implement (de)serialization of the
>> ddm_modify_region. The ddm_modify_region has an extentSet member which
>> is a std::vector<certi:Extent>. But the ddm_modify_region is also a
>> grand daughter of BasicMessage, which have an extent member of the
>> same type. which one should be used? are both required? We are even
>> more confused by the fact that BasicMessage seems to implement
>> serialization of certi::Extent, while serialization of
>> Ddm_Modify_Region and more are (almost) unimplemented. If someone
>> could advise on how we should proceed we could try to make regions
>> work. Thanks.
>
>
> This is a nice proposal. Before giving you appropriate advice I shall
> have a look at that part of the code because I didn't have to look at it
> so often :-).
>
> I'll try to give you advice but this may have to wait until may 25th because
> I'll be on leave next week. In the meantime you may have a look
> at the messaging system of CERTI which is well described in those reports:
> http://download.savannah.gnu.org/releases/certi/papers/Pancik2010_Thesis.pdf
> (in english)
> http://download.savannah.gnu.org/releases/certi/papers/2010-SpecGenMessage_CERTI-RapportDeStage-LucasALBA.pdf
> (in french)
>
> and the presentation:
> http://download.savannah.gnu.org/releases/certi/papers/CERTI-GenMsg-presentation-6june2011.pdf
>
>>
>>
>> /Mattias
>>
>> PS. Also we found that at least 3 functions in DataDistribution.cc
>> that calls setAttributes(...) without having called setAttributesSize
>> before, which causes crashes.
>
>
> I saw your other message and patch; I'll have a look at it soon.
>
> Do not  hesitate to post patch proposal on the tracker:
> https://savannah.nongnu.org/patch/?group=certi
>
>
>>
>> --
>> CERTI-Devel mailing list
>> address@hidden
>> https://lists.nongnu.org/mailman/listinfo/certi-devel
>
>
>
>
> --
> Eric
>
> --
> CERTI-Devel mailing list
> address@hidden
> https://lists.nongnu.org/mailman/listinfo/certi-devel
>