Re: [PATCH v2 00/23] All-stop on top of non-stop

[Pedro Alves <palves@redhat.com>]

On 04/13/2015 05:16 PM, Yao Qi wrote:
> Pedro Alves <palves@redhat.com> writes:
> 
> Pedro,
> I can't apply this patch cleanly, so unable to test it for
> aarch64-linux.  I'll give a test once your V3 is posted, in which this
> patch is included, I assume.

Oh, sorry for not being clearer, I didn't meant for you to try
the patch, only to say that I got a handle in the issue, to avoid you
wasting any time trying to debug it, in case you felt compelled to do
that.  I'll post something testable once I clean things up a bit more.

> 
>> +      /* If we have events pending, go through handle_inferior_event
>> +	 again, picking up a pending event at random.  This avoids
>> +	 thread starvation.  */
>> +      pending = iterate_over_threads (resumed_thread_with_pending_status,
>> +				      NULL);
> 
> A quick question, I don't see how pending events are picked up at
> random.  Do you mean target_ops->to_wait, such as linux_nat_wait, can
> get event at random?
> 

It's done in do_target_wait.  So we go back to the event loop,
and end up in fetch_inferior_event again.  In the series/branch, that
fetch_inferior_event calls do_target_wait instead of target_wait directly:

/* Wrapper for target_wait that first checks whether threads have
   pending status to report before actually asking the target for more
   events.  */

static ptid_t
do_target_wait (ptid_t ptid, struct target_waitstatus *status, int options)
{

which calls random_pending_event_thread, and that is what takes care
of the randomization.

Thanks,
Pedro Alves