Bill, are you proposing to remove calls to
SystemFailure.initiateFailure() and remove all of the emergency-class
stuff or just add a Deprecated flag to SystemFailure?
On 10/28/19 12:01 PM, Bill Burcham wrote:
The SystemFailure class is a clearing house for detecting and attempting to
mitigate SystemFailure exceptions e.g. VirtualMachineError and
OutOfMemoryError.
This class should not exist. SystemFailure exceptions should be allowed to
percolate up and cause the JVM to terminate as soon as possible with an
informative message in the log, if possible.
Here is what the JVM spec has to say [1]:
"A Java Virtual Machine implementation throws an object that is an instance
of a subclass of the class VirtualMethodError (sic) when an internal error
or resource limitation prevents it from implementing the semantics
described in this chapter. This specification cannot predict where internal
errors or resource limitations may be encountered and does not mandate
precisely when they can be reported."
There's a typo in the spec there: it says "VirtualMethodError" when it
means "VirtualMachineError". Anyhoo, the upshot is: the JVM spec does not
apply after you've encountered a VirtualMachineError. As a result, there is
no reason to believe that subsequent processing will make things better
(than exiting immediately).
The SystemFailure class should be deprecated so no new dependencies to it
are added. Existing dependencies on it, should be eliminated over time.
Do you have any objections to deprecating the class and beginning
elimination of usage of it within Geode?
Regards,
Bill
[1] https://docs.oracle.com/javase/specs/jvms/se7/html/jvms-6.html#jvms-6.3
[2] oh and here's a ticket https://issues.apache.org/jira/browse/GEODE-7369
