Oh, I haven't though about that. Thank you for clarifying that to me. 
Meanwhile I rewrote the whole damn thing :)

Thanks again,
Mihai

On Monday 20 October 2003 15:59, Nathan J. Williams wrote:
>
> Additionally, you aren't guaranteed that the new thread will run
> before this thread loops around again and creates more threads - it
> looks like that's what the pth_yield() was trying to do in the
> original code, but trying to use scheduling and luck for
> synchronization doesn't work very well. Since nothing in *this* loop
> touches the file descriptor s, select() will continue to return the
> same status. You need some kind of handoff mechanism so that you only
> create one thread at a time per connection.
>
>         - Nathan