Jenkins 2.303.3 fails to restart itself, e.g. after adjusting plugins (additions, updates, deletions), leaving users to face the “Please wait while Jenkins is restarting” page).

The following trace is left behind:

WARNING jenkins.model.Jenkins$20#run: Failed to restart Jenkins
java.io.IOException: Failed to exec 'jenkins: /usr/bin/java' No such file or directory
  at hudson.lifecycle.UnixLifecycle.restart(UnixLifecycle.java:88)
  at jenkins.model.Jenkins$20.run(Jenkins.java:4482)

Checking the code, there seems to be an absolute trust in the current arguments (except for --daemon getting stripped): https://github.com/jenkinsci/jenkins/blob/stable-2.303/core/src/main/java/hudson/lifecycle/UnixLifecycle.java#L54-L64

That cannot work since daemon version 0.7.0-1 as found in Debian 11+ (published 2021-08), and apparently Ubuntu since Hirsute Hippo (published 2021-04)… prepends the value of the --name parameter in argv[0], contrary to 0.6.4-1 as found in Debian 6-10 and earlier Ubuntu versions.

The relevant code added between 0.6.4-1 and 0.7.0-1: https://sources.debian.org/src/daemon/0.7-1/daemon.c/#L4338-L4351

This can be checked with the ps command (asking for detailed output) or with e.g.:

# tr '\0' '\n' < /proc/<pid>/cmdline
jenkins: /usr/bin/java
-Djava.awt.headless=true
-jar
/usr/share/jenkins/jenkins.war
--webroot=/var/cache/jenkins/war
--httpPort=8080

I'm not sure what a proper fix would look like though. If that kind of fun is limited to newer daemon versions, maybe just rewriting the first argument, dropping a possible leading "jenkins: " string would be sufficient to dodge this issue, without having to rethink the whole restart procedure?