Hi Ludwig,

my setup uses CAF (C++) together with RIOT on a 64 bit Linux (Debian).

I created a branch in my RIOT fork which includes all the code to make the 
example work [1]. You can find it under "test/cpp_timer/“. It depends on an 
implementation for thread, mutex and condition_variable that is mostly STL 
compliant and is based on RIOT. Not STL compliant is, e.g., the chrono 
implementation used in the test … (it is atm. only sufficient to make CAF work).

I added your fixes to the branch. Btw, there are some errors on OS X [2].

Raphael


[1] https://github.com/josephnoir/RIOT/tree/topic/vtimer
[2] https://gist.github.com/josephnoir/a46ce587c87b43a11222

> On Nov 25, 2014, at 6:11 PM, Ludwig Ortmann <[email protected]> 
> wrote:
> 
> Hi,
> 
> any chance for me to get access to the application aka test case?
> Also the specs of your system might be interesting (CPU).
> 
> Cheers, Ludwig
> 
> On Tue, Nov 25, 2014 at 05:56:19PM +0100, Raphael Hiesgen wrote:
>> Hi Ludwig,
>> 
>> this does not change the behavior of my tests:
>> * multiple timers still work
>> * short consecutive ones stop after a while
>> 
>> Raphael
>> 
>> On Mon, 2014-11-24 at 18:57 +0100, Ludwig Ortmann wrote:
>>> Hi Raphael,
>>> 
>>> please test again.
>>> 
>>> The last commit is a kludge, but should at least work for most cases
>>> until I come up with a better solution.
>>> 
>>> Cheers, Ludwig
>>> 
>>> On Mon, Nov 24, 2014 at 02:14:18PM +0100, Ludwig Ortmann wrote:
>>>> Hi,
>>>> 
>>>> Thanks, I'll send a notification when there is further progress.
>>>> 
>>>> Cheers, Ludwig
>>>> 
>>>> Am 24. November 2014 14:06:52 MEZ, schrieb "Hiesgen, Raphael" 
>>>> <[email protected]>:
>>>>> Hello Ludwig,
>>>>> 
>>>>> looks better. Some of the tests that did not work before, work fine
>>>>> now, e.g. using multiple timers.
>>>>> Setting a consecutive timer around 1 milliseconds still seem to end up
>>>>> in a freeze.
>>>>> Most often with the error: "schedule_timer(): timer is already due (3),
>>>>> mitigating.”.
>>>>> 
>>>>> Thanks,
>>>>> Raphael
>>>>> 
>>>>>> On Nov 22, 2014, at 4:08 PM, Ludwig Ortmann
>>>>> <[email protected]> wrote:
>>>>>> 
>>>>>> Hi Raphael,
>>>>>> 
>>>>>> On Wed, Nov 19, 2014 at 06:51:25PM +0100, Ludwig Ortmann wrote:
>>>>>>> On Wed, Nov 19, 2014 at 05:12:42PM +0000, Hiesgen, Raphael wrote:
>>>>>>>> Hey there,
>>>>>>>> 
>>>>>>>> more questions! We encountered some problems using vtimer.  For
>>>>> once, setting multiple short timers consecutively leads to a segfault.
>>>>> We tired this with the timer_msg test reducing the intervals to around
>>>>> 150ms.
>>>>>>>> 
>>>>>>>> A second problem showed up when setting multiple timers at once,
>>>>> e.g., from different threads. After a short time the "timer is already
>>>>> due” warning from schedule timer in hwtimer_cpu.c flashed and soon
>>>>> after the timers simply stopped.
>>>>>>>> 
>>>>>>>> So, I am not sure if these problems originate in our own
>>>>> implementation or the native port. In order to test the short timers,
>>>>> we executed the same program on the iot-lab_M3 board, where it worked
>>>>> even with much shorter timers. 
>>>>>>>> 
>>>>>>>> Anything I should know?
>>>>>>> 
>>>>>>> There is some problem I didn't get around to fixing, yes.
>>>>>>> I guess your issue it is related to this:
>>>>>>> https://github.com/RIOT-OS/RIOT/issues/715
>>>>>> 
>>>>>> Could you try if/how this changes the effect you are seeing in
>>>>> native?
>>>>>> 
>>>>>> https://github.com/RIOT-OS/RIOT/pull/2071
>>>>>> 
>>>>>> I am still able to freeze the timer (especially in Valgrind), but it
>>>>>> does so less often.
>>>>>> 
>>>>>> Cheers, Ludwig
>>>>>> _______________________________________________
>>>>>> devel mailing list
>>>>>> [email protected]
>>>>>> http://lists.riot-os.org/mailman/listinfo/devel
>>>>> 
>>>>> _______________________________________________
>>>>> devel mailing list
>>>>> [email protected]
>>>>> http://lists.riot-os.org/mailman/listinfo/devel
>>>> 
>>>> _______________________________________________
>>>> devel mailing list
>>>> [email protected]
>>>> http://lists.riot-os.org/mailman/listinfo/devel
>>> _______________________________________________
>>> devel mailing list
>>> [email protected]
>>> http://lists.riot-os.org/mailman/listinfo/devel
>> 
>> 
>> _______________________________________________
>> devel mailing list
>> [email protected]
>> http://lists.riot-os.org/mailman/listinfo/devel
> _______________________________________________
> devel mailing list
> [email protected]
> http://lists.riot-os.org/mailman/listinfo/devel

_______________________________________________
devel mailing list
[email protected]
http://lists.riot-os.org/mailman/listinfo/devel

Reply via email to