Re: s6-permafailon not acting as expected

From: Laurent Bercot <>
Date: Wed, 18 Nov 2020 18:35:09 +0000

>Sounds good. So what kind of interface changes should we expect? I
>mean will they require updating the service definitions, run/finish
>scripts, that sort of thing?

  (Sorry, I missed it indeed.)

  Nothing that drastic; service definitions will not change. They are the
core of s6, and such a change would warrant a lifetime version bump,
I think.

  No, the next release will be about a long overdue cleanup of s6-svscan
and s6-svscanctl invocations: command-line options will change, and
default semantics of signals received by s6-svscan will also change.

  The point is to remove some clutter that is either:
  - legacy behaviours from daemontools's svscan that have not shown they
were useful
  - ad-hoc stuff that was added to support early iterations of running
s6-svscan as pid 1, and that isn't used anymore (in part thanks to the
existence of s6-linux-init 1.x)
  - options that experience has shown to be traps (i.e. ideas that seem
good at first but really encourage bad patterns)

and make the s6-svscan/ctl system less confusing to new users by only
keeping a few relevant and useful configuration switches.
  There will also be QoL improvements to s6-svscan for people who like
to nest supervision trees; also, no promises, but I may come up with
some way to officially support user trees, which are a relatively
common request.

Received on Wed Nov 18 2020 - 18:35:09 UTC

This archive was generated by hypermail 2.3.0 : Sun May 09 2021 - 19:44:19 UTC