Most of the problems that I've found so far are actually bugs in init-d-script. The rest are between init-d-script and init-d-script-courier _override() functions not replicating needed functionality.
Before I started looking into this, I had never heard of /lib/init/init-d-script before.
The first thing I did was to find out who else was using it in their init scripts. The answer for me turned out to be almost nobody. Of the nine different hosts I looked at, I could find only one other package which used it: apache2's apache-htcacheclean, for cleaning up after mod_cache_disk. And even there it's not really comparable because of the very limited function it performs. The main apache2 init script doesn't use init-d-script.
And since nobody apparently uses it, it didn't surprise me when I started finding bugs. See bug #822674 for our current most significant issue. Nothing will work until that gets fixed or worked around.
It does not help that PIDFILE is missing from all of the init scripts, except for courier-authdaemon, where it's wrongly declared as PIDFILE="/run/courier/pid". Thus, init-d-script is looking for the PIDFILES in all the wrong locations. This will have to get fixed too, but won't make a difference because of the above mentioned bug.
The "exit 0" at the end of init-d-script is probably also a bug as it destroys almost all return codes. I have like five other issues which might get turned into bugs once I've had a second look at it tomorrow.
Using /etc/init.d/skeleton as a template may have seemed like a reasonable place to start to write a new init script, but I don't know if it's worth working around the issues I've found so far. I'm already using my own init script replacements so things will work.
The Courier suite, being as weird as it is, also makes it a poor candidate to fit into a per-defined structure like init-d-script provides.
Most of our other issues are that functionality normally expected isn't in the init-d-script-courier _override() funcs. There's no start/stop logging, no stdout, etc. That's all fixable though.
I will look at this more later in the week and send you some proposed changes. I'm not sure if init-d-script should be ditched or not yet. I like the idea of a uniform init script, but the issues I've found so far explains why nobody is using it.
On 04/21/2016 02:38 AM, Ondřej Surý wrote:
Please read /lib/init/init-d-script (and /usr/lib/courier/init-d-script-courier) first and then come back again.
