From: Robert Feldt <feldt@...>
Date: 2000-10-27T22:59:34+09:00
Subject: [ruby-talk:5915] Re: [RFC] Towards a new synchronisation primitive

On Fri, 27 Oct 2000, hipster wrote:

> We need a careful definition of the Ruby scheduler granularity here.
> Otherwise even an expression like
> 
> 	if test_and_set(bar) ...
> 
> can not be guaranteed to be atomic: reschedule between test_and_set
> and if-evaluation. (or is this impossible?)
> 
You're right but I was thinking along the lines:

def test_and_set(ref_to_bool_var)
	Thread.critical = true
	previous_value = ref_to_bool_var
	ref_to_bool_var = true
	Thread.critical = false
	previous_value
end	

only that it is implemented in C and we dont "export" Thread.critical= (I
think it's rb_thread_critical_set) to language level (or not even to C
level I guess...).

Regards,

Robert