> If I understand you correctly, this could be done by inserting `/pat/ { print
> | "post-process" }` awk script/`forstdin`/etc. ) in a pipeline after `s6-log
> ... 1`. Either s6-rc's pipeline mechanism, or redirecting s6-log's stdout to
> a plain ol' named pipe if you don't use the former.
Thanks, that's also a solution, so the pipeline will look like
`producer | s6-log | post-processor`,
so it can be done already with what it there.
>
> Sometimes what I miss in s6-log is having something like `1` but for an
> arbitrary preopened file descriptor, so lines can be selectively
> dispatched/copied to multiple pipelines, but I haven't needed such
> functionality yet.
>
It might be useful for example when you use s6 to supervise a
multi-process application
(so s6 is used in this case not as an init system, but rather as a
generic supervisors for a
collection of programs). For example for computational biology when you need to
process a lot of data it would be cool to be able to create supervised
pipelines with a
more complex topology with splits and merges.
--
Alex
