Re: [PATCH] Fix PR18360 - internal error when using "interrupt -a"

[Pedro Alves <palves at redhat dot com>]

On 11/12/2016 12:49 AM, Luis Machado wrote:

>> @@ -6051,6 +6032,15 @@ handle_signal_stop (struct
>> execution_control_state *ecs)
>>    if (random_signal)
>>      random_signal = !stopped_by_watchpoint;
>>
>> +  /* Always stop if the user explicitly requested this thread to
>> +     remain stopped.  */
>> +  if (ecs->event_thread->stop_requested)
>> +    {
>> +      random_signal = 1;
>> +      if (debug_infrun)
>> +    fprintf_unfiltered (gdb_stdlog, "infrun: user-requested stop\n");
>> +    }
>> +
> 
> Should we output information about the event thread here to make this
> message a bit more meaningful and make debugging problems easier?

It's already printed earlier, I believe.  Most tracing debug logs
don't print the thread info, which may actually make comparing logs
with diffing tools easier.

> The test cookbook says we should avoid having infinite loops in case GDB
> crashes. Can we change this so we have something we can expect to finish?

Actually, the test already addresses this by following option 2
suggested in the cookbook: "Use alarm.".  It's in main:

+  alarm (300);

> 
> Otherwise i have no other comments on this.

Thanks!

Meanwhile, I managed to trip the new assertions put in place to ensure
that GDB doesn't resume threads when they should be help stopped.  We'll
need a further tweak.  I'll post a v2 soon.

-- 
Pedro Alves