------ Original Message ------
From "Carlos Eduardo" <carana2...@gmail.com>
To supervis...@list.skarnet.org Cc "Carlos Eduardo" <carana2...@gmail.com> Date 2024-05-04 21:58:21 Subject [PATCH] Add adoption to trap
With this patch, trap(1) is able to accept an existing PID as prog..., if given the new -P option. I'm sending this as patch because the only scenario where this does something that isn't feasible with execline's current wait(1) command or the original interface to trap is if you use Linux or Free/DragonFly BSD's subreaper functionality to reuse trap as a turbo-fghack that forwards signals and tracks the exit code of the forked-off daemon instead of the short-lived daemonizer.
Is there a reason why you're sending your s6 reports to the skaware ML and your execline patches to the supervision one? 🤨 I'm not going to apply that patch, because as you said, it does not makes sense outside of subreapers, and I consider subreapers harmful (outside of a new pid namespace, which requires other kernel work and surrounding setup so it's a special case anyway, there's nothing significant that can be achieved with them that cannot be achieved without them; they're not portable and they make process hierarchy more confusing, so the trade-off is negative). People who are going to mess around with subreapers and SIGCHLD redirection hacks are more than capable of building their own stuff to work with it without involving execline's trap. -- Laurent
