It's (unfortunately?) normal that a JVM doesn't exit on OOME. If you
have a top-level processing loop, you can do a:
try
{
...
}
catch(VirtualMachineError e)
{
System.exit(123);
}
or similar. I tend to have such catches in my production stuff, since
if you get a VirtualMachineError (of which OOME is one subclass), then
all bets are off with regard to correct future execution of that VM
instance -- you can't count even on finally blocks completing, streams
being flushed (potentially leading to incomplete file writes), etc.,
so indeed the best course of action is exiting as quickly as possible
and making sure monitoring staff spots it immediately.
One possible exception you might want to make is for
StackOverflowError -- it too is a VirtualMachineError, but is normally
localized in effect to a single thread. However, it too can cause VM
guarantees to fail; a finally block executing after a SOE might get
another SOE itself if the stack is still very deep while in the
finally block, so I usually don't bother making an exception for it
and my code does a System.exit() whenever it spots any
VirtualMachineError.
Attila.
--
home: http://www.szegedi.org
twitter: http://twitter.com/szegedi
weblog: http://constc.blogspot.com
On 2009.01.09., at 16:10, Tom Robinson wrote:
I have a fairly complex JavaScript app I run using the Rhino shell.
Occasionally it runs out of memory and throws the OutOfMemoryError
exception. I expect the java process to exit at that point, but it
doesn't. No code seems to be executed after that point, but the
process doesn't exit.
This is rather annoying because then I can't check to see if the
process has crashed and automatically restart it.
I'm having a hard time creating a simple reduction. Has anyone
experienced anything like this in Rhino, or Java in general?
Thanks.
_______________________________________________
dev-tech-js-engine-rhino mailing list
[email protected]
https://lists.mozilla.org/listinfo/dev-tech-js-engine-rhino
