Antw: Re: [lwip-users] SemSignal gets called on a already signaled semaphore

[Matthias Weisser]

>>> Jonathan Larmour <address@hidden> schrieb am 31.05.2007 um 11:20:
> Matthias Weisser wrote:
>> Hi
>> 
>> After playing around with my echo server I saw a strange behavior of my
>> sys_sem_signal() implementation. I have implemented the semaphores as
>> binary mutexes. Now I see that one of the semaphores is created with a
>> initial count of 0 then signaled and then signaled again. Can this be
>> ignored as I expect or indicates it another bug in my port?
> 
> Semaphores and mutexes are not identical. Especially, as in this case, when 
> the semaphores are counting semaphores, not binary. But you can abstract it 
> using mutexes and condition variables, and a variable to represent the 
> semaphore count.

So, this means that lwip needs counting semaphores? If so than I can use the 
semaphores of my OS (which are counting semaphores) but than I have still the 
problem of a lot of sys_sem_signal calls to a semaphore where is never called a 
wait for. So the semaphores value is incremented and incremented and 
incremented until it reaches a maximum value (which I can define in my OS) and 
then the OS reports the error. I can set the maximum to a very high value but 
that doesn't solve the problem.

Another question is: What does the count argument of sys_sem_new mean? Is it 
the initial count value of the semaphore? Or anything else?

Matthias