Logging JVM exit (in context of JEP 486)

[Michael Bien]

On 10.10.24 11:16, Alan Bateman wrote:
>
>
> On 10/10/2024 08:57, Michael Bien wrote:
>> Hello,
>>
>> Calls to System.exit() can be currently tracked by setting the 
>> "java.lang.Runtime" logger to DEBUG (it even dumps the stack trace 
>> which is nice). However, I couldn't find an equivalent for 
>> Runtime.getRuntime().halt(). Are there plans to add more logging to 
>> debug common JVM exit events?
>>
>> (if this was already discussed elsewhere - super sorry wasn't able to 
>> find anything)
>
> Logging involves executing arbitrary Java so not for halt time. It's 
> come up a few times, I think the most recent on core-libs-dev was Aug 
> and/or Sept 2023, you'll need to search through the archives to find it.

Thanks! I found JDK-8301627 and JDK-8303198 and the discussion on the 
PRs also gave some context.

+ https://www.mail-archive.com/core-libs-dev@openjdk.org/msg19309.html

(I made the mistake to mainly search through security manager threads 
since I somehow expected that it would be discussed there too)


The JVM would still be able to log something to console (in native 
beforeHalt() or halt0()) without the application/logging callback issue?

-Xlog:safepoint+stats=info seems to log some kind of exit cause, it is 
the wrong way around though :)

System.exit() / halt() will print:
[0,419s][info][safepoint,stats] Exit       1

regular process exit (no-non daemon threads remaining) prints:

[0,454s][info][safepoint,stats] Halt      1


best regards,

michael

>
> -Alan