Troubles with Shenandoah

[Aleksey Shipilev]

On 4/9/19 11:09 PM, Simone Bordet wrote:
> Ran 12+33 with -Xms==-Xmx and -XX+AlwaysPreTouch for 20+ mins without issues.
> 
> Now I'm just curious about what the previous settings triggered.

My bet is on long commits after full uncommit triggered by System.gc() wrecked up everything. Maybe
it is not a GC bug, but rather application layer loses the messages with timeouts? It would be a bit
better for your -Xms16g (if I recall correctly?) setting after this lands:
  https://bugs.openjdk.java.net/browse/JDK-8222186

...but -Xms = -Xmx and AlwaysPreTouch is also good.

> How come https://bugs.openjdk.java.net/browse/JDK-8222129 is not
> triggered with this setup, but was with the previous?

This failure is about the feature (load-reference-barriers) not backported to 12u yet. Either way,
it is not fatal, and release build would just take a safe default.

-Aleksey