[En-Nut-Discussion] reliability of message queues

[Harald Kipp]

>
>some thoughts: the message queue is using a lot of critical sections.
>if you don't need to send messages from an interrupt handler, I would
>advise to write a new version which doesn't use any critical sections
>(which are bad for bt uart..:)

I agree with this view.


>does anybody NEED to send msgs from an IRQ handler?

Probably Ralph Mason, who contributed it. I'd suggest
to create an alternative module msgqueue.c and implement
NutMsgQueueXXX() instead of NutMsgQXXX().


Harald

P.S.: IMHO, Nut/OS kernel routines should stick with the basic
NutEvent API for any kind of thread synchronization. But for
higher level APIs and application code message queues and
semaphores are quite useful.