[runwsgi] Cannot override [DEFAULT] section entries from the command line

[dwt]

What I did:

Given a zope config like this:

[DEFAULT]
http_port = 8000

[app:zope]
use = egg:Zope#main
zope_conf = %(here)s/zope.conf

[pipeline:main]
pipeline =
    egg:Zope#httpexceptions
    zope

[server:main]
use = egg:waitress#main
host = 127.0.0.1
port = %(http_port)s

I tried overriding that value from the command line by calling runwsgi like this

runwsgi --verbose --debug instances/test/etc/wsgi.ini http_port=8001

I expected the resulting server to run on port 8001 - but it didn't and instead ran on DEFAULT.http_port.

What I expect to happen:

There should be a way to verride values from the command line while allowing a default value from the configuration file.

What actually happened:

Overriding not possible.

What version of Python and Zope/Addons I am using:

% python --version                                                                                                                            :(
Python 3.6.8
% pip list |grep -i zope
Zope                             5.1.1.dev0 /Users/dwt/Code/Projekte/zope4/Zope/src

[icemac]

I am not sure if your expectation should be matched by a code change as it makes the code more complex for a use case which seems exotic to me. But let's see what others think.

[dwt]

Well, I would have expected that command-line values have more precedence than config file options.

For development at least I would like to quickly override a setting from the shell to test out the effect. Also something like this doesn't work: runwsgi --verbose --debug instances/test/etc/wsgi.ini server:main.port=8001 to force a different port.

The workaround I used is to have a non defaulted http_port variable in the config, that is then provided on the command line, but that means that one has to always provide it. That is easily worked around by a wrapper script which sets it if not provided - but it sucks.

I ran into this when I tried to run multiple workers for a zeo deployment and tried to instantiate them from a systemd template unit like this:

# {{ ansible_managed }}
[Unit]
Description=ZMS/Zope production environment
After=network-online.target
PartOf=zms-{{ zope_instance.name }}.target
After=zms-{{ zope_instance.name }}.target

[Service]
Type=simple
User=zope
Restart=always
Environment=HTTP_PORT=%i
ExecStart=/home/zope/instances/{{ zope_instance.name }}/bin/runzope

[Install]
WantedBy=zms-{{ zope_instance.name }}.target

This is then instantiated by enabling systemd services like this: `systemd enable --now zms-$instance_name.zope@{8001..8006}.service

Right now, this hands in the port via an environment variable and then calls a wrapper script that defaults that port to ensure that for debugging, the zope service can still be easily started on the default port manually. I thought it more intuitive if this was achievable by directly calling runwsgi for example.