Re: [pok-devel] Join the POK team

[Etienne Borde <etienne.borde@xxxxxxxxxxxxxxxxxxxx>]

Hi Julien,

Thanks for your answer. Antoine is not yet registered to the mailing list; I thus transfer this email to him so that he can answer to you.

Best regards,
Etienne.

----- Mail original -----
De: "Julien Delange" <julien@xxxxxxxxx>
À: pok-devel@xxxxxxxxxxxxxxxxxxx
Envoyé: Jeudi 13 Février 2014 15:17:31
Objet: Re: [pok-devel] Join the POK team




Hi Antoine,

The best approach would be to add you to the committers, right? Can you register on tuxfamily, send me by e-mail your username so that I can add you to the project and you can commit your changes.

Would that sound ok?




On Thu, Feb 13, 2014 at 2:16 AM, Antoine Jaouen < antoine.jaouen@xxxxxxxxxxxxxx > wrote:







Hello,



I am, Antoine Jaouën, a PhD student under the direction of Laurent Pautet, Etienne Borde and Thomas Robert from the LTCI - Telecom ParisTech.



I would like to join your team in order to integrate my work into your project POK and help your team to improve the project.



I developed a deterministic implementation of ARINC653 inter-partition communications based on the Periodic-Delayed protocol introduced in : Deterministic Implementation of Periodic-Delayed Communications and Experimentation in AADL. Fabien Cadoret, Thomas Robert, Etienne Borde, Laurent Pautet, and Frank Singhoff. ISORC 2013.




In order to start a collaboration, I want to share with you some bugs I found in the current version of POK. The first bug is in the next_activation time of threads after a call of pok_thread_period (which induces a thread to wait until its next period time).



The bug is located in kernel/core/sched.c :

line 206 :
thread->next_activation = thread->period + POK_GETTICK();



This way, a thread waits its full period duration after ending its execution. Nevertheless it must wait only the remaining time until its next period time.



Here is the correction : ​
thread->next_activation = thread->next_activation + thread->period;




I also corrected the toolchain. The path of 'ar' was hard-coded in 'misc/mk/ rules-partition.mk '.


My last correction was the allocation of 'Partition_Slots' in 'examples/arinc653-queueing/model.aadl', which were too long. I have not checked/corrected all examples but some others have also issues with regard to the dimension of 'Major_Frame', 'Partition_Slots' and 'Period' of threads.





You will find a patch as an attachment of this email, which corrects all those issues.





Best regards,

​Antoine Jaouën​













Antoine JAOUËN
Doctorant / PhD. Student
Institut de Recherche Technologique SystemX
8, avenue de la Vauve - 91120 PALAISEAU
Phone +33 1 69 08 05 60
Site web | Twitter | LinkedIn