RFR: Make sure debug builds fail predictably on safepoint timeout

[Zhengyu Gu]

Okay.

-Zhengyu

On 10/01/2018 02:26 PM, Aleksey Shipilev wrote:
> In sh/jdk8, we have a very intermittent deadlock on VM shutdown in stress tests. gdb session says it
> is caused by safepoint synchronization never completing. Let's make sure we are failing reliably
> when that happens:
> 
> diff -r cb6ddc702b08 src/hotspot/share/gc/shenandoah/shenandoahArguments.cpp
> --- a/src/hotspot/share/gc/shenandoah/shenandoahArguments.cpp   Mon Oct 01 19:11:33 2018 +0200
> +++ b/src/hotspot/share/gc/shenandoah/shenandoahArguments.cpp   Mon Oct 01 20:23:11 2018 +0200
> @@ -192,6 +192,16 @@
>       FLAG_SET_DEFAULT(NodeLimitFudgeFactor, NodeLimitFudgeFactor * 3);
>     }
>   #endif
> +
> +  // Make sure safepoint deadlocks are failing predictably. This sets up VM to report
> +  // fatal error after 10 seconds of wait for safepoint syncronization (not the VM
> +  // operation itself). There is no good reason why Shenandoah would spend that
> +  // much time synchronizing.
> +#ifdef ASSERT
> +  FLAG_SET_DEFAULT(SafepointTimeout, true);
> +  FLAG_SET_DEFAULT(SafepointTimeoutDelay, 10000);
> +  FLAG_SET_DEFAULT(DieOnSafepointTimeout, true);
> +#endif
>   }
> 
>   size_t ShenandoahArguments::conservative_max_heap_alignment() {
> 
> Testing: tier3_gc_shenandoah
> 
> Thanks,
> -Aleksey
>