[omniORB] Condition varaibles - follow up...

[Ole Storm]

Hi Shinji,

Shinji Suzuki wrote:
> (...)
> you may have to do
> 
> BufferItem
> Buffer_i::GetItem()
> {
>    omni_mutex_lock l(omni_mutex_cond_mutex); // safer this way than m.lock()
>    while( !index ) {
>     // No items awailable. Wait on the condition varaible empty_buf:
>      empty_buf.wait();
>    }
>    BufferItem itm = TheBuffer[--index];
>    cout << "Returned item: " << item << "\n" << flush
>    // Signal one of the possible waiting threads on PutItem()
>    full_buf.signal();
>    return itm;
> }
> 
> and apply similar changes elsewhere.

I agree that the omni_mutex_lock class is a better way to mutex-lock an
_entire_ function, but I don't think that this has anything to do with the
problems I experience on WinNT. The example works fine on Linux without
using the omni_mutex_lock - remember.

Actually, mutex-locking the entire GetItem() and PutItem() functions as
proposed will result in a dead-lock at some point. If, for instance
GetItem() is called on an empty buffer, the call will be set to wait in the
call to empty_buf.wait(). But since GetItem() locks on the global mutex
variable, PutItem() is never allowed to execute --> no items will be put
into the buffer --> the thread waiting for new items will never be signalled
--> dead-lock.
This was the reason for calling mutext_region.lock() _after_ the while loop
and not before it!


Thanks anyway and best regards,

	Ole.

---------------------------------------------------------------
Ole Storm
The Institute of Applied Computer Science (IFAD)
Forskerparken 10, DK-5230 Odense M, Denmark
Phone: +45 6315 7134, Fax: +45 6593 2999, Email: storm@ifad.dk
WWW: http://www.ifad.dk
---------------------------------------------------------------