[devUtils/procRunner] wait for proc to exit so we fallback to SIGKILL

[spalger]

When the ProcRunner tries to kill a process, but the process never exits, it looks like it's going to try and send a SIGKILL to avoid the Error: Proc "kibana" was stopped but never emitted either the "exit" or "error" event after 30000 ms errors we see in CI every once and a while, but since it's not waiting for the outcome promise it isn't ever reaching that bit of logic.

I think I wrote this originally, and I also think I assumed treeKillAsync() wouldn't resolve until all the processes had exited, but it resolves the promise as soon as it has sent the signal and doesn't care about the result of the process.

[elasticmachine]

💚 Build Succeeded

• continuous-integration/kibana-ci/pull-request

[tylersmalley]

Not sure if it's something you want to address - but since we don't wait for the outcome, we don't know what the exit code is:

info [kibana] exited with null after an hour

[spalger]

@tylersmalley when we kill the process with using a signal we get null back as the exit code. That log line is produced specifically by observing the proc.outcome$ so it should be shown if it could be known.

[spalger]

6.x/6.4: 1bce3b4