# Settings for svctl(8).
#
# SVCTL_TIMEOUT -- time limit in milliseconds for every s6-rc state transition
# svctl triggers: "svctl update", "svctl start", "svctl stop",
# "svctl restart". 0 means "no limit" and is s6-rc's own default; that is
# exactly what made the postinst of s6-policy-exclusive hang for more than ten
# minutes on rt-2.tick1 on 22.09.2026, because a oneshot did not come up --
# dpkg held its lock, the package stayed in "iF", and every further package
# operation on the machine stood still.
#
# 120000 (two minutes) is generous for anything that happens while packaging,
# and short enough that a hanging service produces a message instead of a
# stalled machine. Whoever has a service that needs longer raises the value
# here -- or sets it to 0 and accepts the old behaviour deliberately.
SVCTL_TIMEOUT=120000

# SVCTL_TIMEOUT_DPKG -- the same, but for transitions INSIDE a dpkg run
# (detected by DPKG_RUNNING_VERSION, which dpkg sets in every maintainer
# script).
#
# Why shorter: while dpkg runs, every wait holds the dpkg lock, and "the
# service does not come up on first install" is the normal case there -- its
# configuration is still missing. Debian's zabbix-proxy-sqlite3.postinst starts
# the proxy twice and stops it before; with 120000 ms that turned into an
# installation of several minutes in which no other package on the machine
# could do anything (measured 23.09.2026).
#
# Outside dpkg SVCTL_TIMEOUT applies: there a human or Puppet waits for a
# service that is really supposed to come up.
SVCTL_TIMEOUT_DPKG=20000
