We have a regression on LP:#1554863 in 2.6-beta1 (installed: 2.6-beta1+develop-4ad365b (6538) 71MB classic):
c@slate:~/Projects/go/omnibus/src/github.com/CanonicalLtd/omnibus$ juju bootstrap localhost --config metering_url=http://40.113.217.209/v3
WARNING unknown config field "metering_url"
Creating Juju controller "localhost-localhost" on localhost/localhost
Looking for packaged Juju agent version 2.6-beta1 for amd64
No packaged binary found, preparing local Juju agent binary
WARNING unknown config field "metering_url"
To configure your system to better support LXD containers, please see: https://github.com/lxc/lxd/blob/master/doc/production-setup.md
Launching controller instance(s) on localhost/localhost...
- juju-18e59d-0 (arch=amd64)
Installing Juju agent on bootstrap instance
Fetching Juju GUI 2.14.0
Waiting for address
Attempting to connect to 10.130.222.88:22
Connected to 10.130.222.88
Running machine configuration script...
Bootstrap agent now started
Contacting Juju controller at 10.130.222.88 to verify accessibility...
Bootstrap complete, "localhost-localhost" controller now available
Controller machines are in the "controller" model
Initial model "default" added
Did Juju ever error as opposed to warn about config attributes that didn't match the schema? For a long time, Juju had no way of allowing model config that was provider specific, so it had to allow unknown attributes to be set - the provider would use them but the attribute was unrecognised by Juju's config schema. eg AWS clouds can have a vpc-id, Openstack can have use-floating-ip etc. So Juju just used to warn about such values. And I suspect that behaviour still exists today. Or are you saying for controller config, we used to be more strict and genuinely did error?