Re: priorities of newly-created threads

More specifically, this is the patch to thread_create_core I propose.

	David

--- thread.c~   2007-07-24 23:30:39.000000000 -0700
+++ thread.c    2007-08-09 22:11:46.000000000 -0700
@@ -350,6 +350,8 @@
      th->first_func = fn;
      th->first_func_arg = arg;

+    th->priority = GET_THREAD()->priority;
+
      native_mutex_initialize(&th->interrupt_lock);
      /* kick thread */
      st_insert(th->vm->living_threads, thval, (st_data_t) th->thread_id);


David Flanagan wrote:
> Nobu,
> 
> I didn't know that about threads in Linux.  Looks like Java has had the 
> same limitation for years, and I never knew it!  Thanks for the 
> explanation.
> 
> Still, I think that you ought to pretend that threads have priorities, 
> for the sake of API compatibility, even though they are meaningless.
> 
> Thread.current.priority = 2  # Allowed, but does nothing on Linux
> Thread.current.priority           # => 2: because we're pretending
> t = Thread.new { sleep }.priority # => 0: I think it should be 2
> 
> Even though it is meaningless, I think that last line of code should 
> return 2.  Or the second line of code should always return 0.
> 
>     David
> 
> Nobuyoshi Nakada wrote:
>> Hi,
>>
>> At Thu, 9 Aug 2007 08:38:41 +0900,
>> David Flanagan wrote in [ruby-core:11876]:
>>> In Ruby 1.9, newly created threads start at priority 0.  I suspect 
>>> that this is a bug.
>>
>> The current VM uses native threads.  On linux, the scheduling
>> policy is defaulted to SCHED_OTHER, and the priority is bound
>> on 0 in that case.  Therefore, it's rather the restriction of
>> the platform.
>>
> 
>