upgrader startup failure with local provider

[Dimiter Naydenov]

http://paste.ubuntu.com/6159055/

Steps to reproduce:
1. Use trunk (my revision is currently 1888)
2. Make sure do go install . in cmd/juju/ and cmd/jujud/
3. Bootstrap a local environment with --upload-tools and the default boilerplate config (nothing else than admin-secret and type).
4. After the bootstrap is done, try juju deploy anything (wordpress will do)
5. Look in ~/.juju/local/log/all-machines.log to see the error as pasted above.

[Dimiter Naydenov (dimitern)]

After investigating, it turns out the regression was introduced in r1877. Still investigating the exact cause.

[Dimiter Naydenov (dimitern)]

So in r1877 the upgrader error started to appear. But the other error:

ERROR juju.provider.local environ.go:180 failed to create new environ config: storage-port: expected int, got 8040

was not there, so I'll continue to investigate further Paste of my machine-0.log from this run:
http://paste.ubuntu.com/6159692/

[Dimiter Naydenov (dimitern)]

The other 3 errors:

2013-09-26 18:04:58 ERROR juju.provider.local environ.go:180 failed to create new environ config: storage-port: expected int, got 8040
2013-09-26 18:04:58 ERROR juju.provider.local environprovider.go:48 failure setting config: storage-port: expected int, got 8040
2013-09-26 18:04:58 ERROR juju environ.go:48 worker: loaded invalid environment configuration: storage-port: expected int, got 8040

start to show in r1884 (one of mine, when we switch the provisioner to the API). Apparently somehow the "storage-port" environment setting gets coerced from int to string. Not sure why yet, still investigating. We don't change the environment config in any way when we return it from ProvisionerAPI.EnvironConfig() (it's map[string]interface{}, as config.Config.AllAttrs() returns).

[Tim Penhey (thumper)]

Due to changes in how the config was passed, it is now going through json, which changes the int to a float64.

APIPort uses schema.ForceInt

The local provider was just using schema.Int.