Re: [rvm-research] Thread's join issue

[Filip P. <pi...@pu...>]

Hi,

To my understanding, your program will not terminate until join() gets 
called on t1, t2, and t3.  The question is then, whether join() gets 
called before the t[i] thread completes (thereby causing MainThread to 
block), or after t[i] completes (thereby causing RVMThread.join() to 
return immediately).

You're absolutely right that if t[i] has already terminated, its thread 
ID's (as in the system's TID, the C library's pthread_t, and RVM's 
threadSlot) will have likely been relinquished.  That's just the nature 
of the beast; there's nothing you can do about it.

I am confused about your question though - *every* call to Thread.join() 
should result in a call to RVMThread.join().  Thread.join() always calls 
VMThread.join(), which always calls RVMThread.join().  There is no other 
path that the code could ever take.

If you want to store info about a thread so that RVMThread.join() sees 
it, you should be able to do that just fine - just put it into the 
RVMThread class as an instance field.

-Filip



On 1/18/2010 6:42 PM, Xinwei Xie wrote:
> Hi all,
> I've got such question about join problem in RVMThread.
>
> MainThread creates 3 threads: t1, t2, t3 and delegates some task to them to
> process. Then MainThread calls
> for(int i=0; i<3; i++)
> t[i].join();
>
> The question is, it seems that t[i] may terminate before MainThread calls
> t[i]'s join as t[i] finishes its task earlier. In this case, MainThread
> won't *actually* call RVMThread#join, will it? If I want to retrieve some
> information such as its tid and print out at the time of MainThread calls
> t[i]'s join, more specifically, I put my own method in RVMThread#join, how
> can I do that. I observe that RVMThread#join won't be called with respective
> to the *dead* thread. Therefore the *dead* thread's information is kind of
> missing.
>
> Is there anyone who is familiar with this issue can answer. Thanks a
> million!
>
> Regards,
> Xinwei
>