Let SimpleCommandLauncher kill the agent JVM more promptly when disconnected

[jglick]

jenkinsci/support-core-plugin#624 (comment)

[dwnusbaum]

Seems fine. Do you think it is worth running through the PCT to check if it causes any problems for agent disconnection/restart related tests?

[jglick]

Do you think it is worth running through the PCT

That does sound prudent. I will have to look up how to test a JTH bump in PCT.

[jglick]

Turns out it is pretty complex due to a history of core changes: jenkinsci/plugin-compat-tester#746

[timja]

Turns out it is pretty complex due to a history of core changes: jenkinsci/plugin-compat-tester#746

yeah you need to backport this change to the pre jetty 12 line of the test harness and then make a PR to plugin-compat-tester updating the minimum versions of JTH on both lines, then PR to bom with the updated PCT version

[jglick]

I think I am not going to bother with all that. Will try this in a couple of plugins, and if later it turns out to cause any issues, we can just revert/amend.

[jglick]

Breaks ExecutorStepTest.nodeDisconnectMissingContextVariableException though I am not yet sure why. Also does not seem to suffice in every single case in support-core or in workflow-cps, though again I am unsure why. I was able to confirm that this does kill the agent process before JenkinsRule test cleanup begins, so what else is holding the file handle? Unfortunately Windows file lock error messages do not give details, and I have been unable to reproduce any of these failures locally.

[jglick]

jenkinsci/support-core-plugin#624 still not stable. I think I actually meant to use afterDisconnect.