"Laurent Bercot" <ska-supervis...@skarnet.org> writes:
You do not need a shebang in an up script. Up scripts are not
executable files. If you change the interpreter in the shebang,
the
up script will *still* be processed by execlineb.
i feel like this might possibly be a reason why a number of people
assume that _all_ s6-ecosystem scripts have to be execline
scripts. A `run` script in an s6 servicedir can be any executable
file, but an s6-rc `up` script for a oneshot (which i think many
people might feel is roughly similar to a `run` script, in that it
specifies a particular program to run) is always parsed by
execlineb. Of course, that doesn't mean that one can't call some
other interpreter from the execline script, as you note in the
s6-rc-compile documentation:
They can be written in any scripting language by invoking the
interpreter directly: for instance ‘/bin/sh -c “script”’, where
script is a shell script.
However, while it's true that in both s6 and s6-rc, use of
execline is not _ultimately_ required, in the s6-rc `up` context,
one _does_ have to at least use execline to call the preferred
interpreter with the relevant script.
Alexis.
