yosifkit

Official-images should not be using a process supervisor. It should be "one process" per container; this keeps things consistent, composable, and (hopefully) scalable. Similar, now closed issue of adding a...

We did a lot of refactoring and "functionalizing" in https://github.com/docker-library/postgres/pull/496; combine that with https://github.com/docker-library/postgres/pull/452 and it should be possible to call any function you want from the orginial entrypoint. The...

As just an interested party, I would vote against this addition. `tini`'s strength is its limited scope and this seems outside that. Restarting stopped containers should be done by whatever...

I think this is "out of scope" for tini: https://github.com/krallin/tini/issues/71#issuecomment-272652178, https://github.com/krallin/tini/pull/146

It's likely not tini's fault; stdout/stderr ordering is hard. Like this issue in Docker itself: ~https://github.com/moby/moby/issues/26986~ https://github.com/moby/moby/issues/31706

Use `tini` to start `systemd`? Tini won't really be adding anything useful, just running systemd directly will provide zombie reaping. (Can systemd even run correctly when not pid 1?)

As far as I understand it, your problem is in trying to run more than one thing. Tini can only reap processes that have no parent (ie are re-parented to...

I was just thinking that it might still be possible to get a zombie: - say the entrypoint is pid 6 - it spawns your new daemon script to run...

I think this is "out of scope" for tini: https://github.com/krallin/tini/issues/71#issuecomment-272652178, https://github.com/krallin/tini/pull/146, https://github.com/krallin/tini/pull/187

`tini` is only designed to wait for its direct child to exit; if that child spawned its own children, then it should be signaling/waiting for them before exiting. `tini` will...