Jan Kiszka wrote:
Hi,

today two issue hit us, one with RTAI 3.1 and one with magma. First the more critical one:

Paolo, do you remember our task deletion problem with the lxrt scheduler in 3.1? It was this pattern that triggered it, we use it in RTnet for instance:

task_fun()
{
    while (1) {
        if (rt_sem_wait(&sem) == SEM_ERR)
            return;
        /* do some stuff */
    }
}

init()
{
    rt_typed_sem_init(&sem, 0, CNT_SEM);
    rt_task_init(&task, task_fun, ...);
}

cleanup()
{
    rt_sem_delete(&sem);
    rt_task_delete(&task);
}

It still persists in 3.1's CVS, while it is not existing in the 3.1 release. The point is that we need the CVS due to some other bug in the release scheduler, but we can't use it with RTnet because of this bug. Is this issue going to be addressed before 3.1r1?

We have also examined if magma behaves better in both regards. Fortunately, it does! Ok, switching to magma might be an option.


I'm not happy of the magma solution but to have it better I need time and still there is something else to do it will be left pending. As for 3.1 (vesuvio) I must say that I find it difficult to maintain, amid the few things I've to do already. So I'm caring much of 3.2 which is as stable as 3.1, or even better. See ahead.


At this chance, I wanted to enable magma detection and support in RTnet's build system. But instead of some version string like "3.2-testX", "rtai-config --version" just returns "magma"! We all learned now that magma will always be the name of the current experimental RTAI branch, right? So, how shall we distinguish in RTnet between current magma and any potential future magmas?

What about a naming scheme like "3.2-magma" or "magma-post3.1". I don't know, pick any, but please make it distinguishable!


It should be solved now. Philippe already slapped my face because of magma misuse, as I should have set a new stable CVS for 3.2 as soon as magma was working to the point of releasing test versions.

So magma is of no use for anybody now, but people thingking to 3.3. The official CVS for 3.2 is "vesuvio" now. Philippe has set it up according to adhere to his published RTAI development scheme. I cannot just apologise for my careless.

I think your problem should be solved by "vulcano". It is there I'll put your fixes too.

BTW: "vulcano" is not a generic name but the name of Vulcano island, which has a vulcano that erupted last in 1888-1890. There is still some vulcanic activity there that is monitored continuosly by researchers. If I'm not wrong it should be where god Vulcano had his forge.

Paolo.

Paolo.


Reply via email to