Rather than always assuming a missing `WORKDIR` means herokuish, we instead inspect the image to verify that it is. If it is, then and _only_ then do we set WORKDIR. Otherwise, we respect the decision of `docker cp` to execute from within the last known `WORKDIR` context, which is inherited at the image level, not image metadata level.
Additionally, speed up `is_image_herokuish_based` by inspecting the environment variables on the image. When there is a "USER=herokuishuser", we can more or less safely assume it is an image that Dokku built, and is therefore a herokuish image. While possible, it would be very strange if a non-herokuish image had this environment variable set, so it is a relatively safe change.
Previously, if running 'dokku report --all', the early exit 1 in apps:report would cause the entire process to bail, resulting in a partial report. Instead, we exit 0, allowing plugn to continue.
Technically a non-deployed app is not an error case at this position.
While I do not agree with _every_ style change, this will force Dokku to have consistent formatting across all shell scripts, which is arguably a Good Thing™.
The command used to reprocess everything is:
```shell
shfmt -l -bn -ci -i 2 -w .
```
This will allow us to _actually_ introspect on the running state of an application. In certain cases, the detected state can be "mixed", meaning that we may need to take action, depending on what we want to do in mixed mode.
This is useful in cases where the app url has some "interesting" structure - such as a subdomain or scheme change - for a specific installation.
Closes#3160
As this was not strictly announced in our migration guide, removing it is a bc-incompatible change we are not comfortable with. We may wish to remove it in the future
Unlocking an app will not stop running deploys, so users may need to clean up. However, it is useful for deploys that fail in unusual ways that cause the lock to stay around.
Closes#2883
These should never have been in common, and more directly relate to domain handling than anything else.
As this is a BC-break, merging will need to wait until the next minor.
- use a plugin trigger to see whether we should bind to all interfaces
- create a generic way of setting properties for a plugin
- migrate proxy-enabled to the new network property "bind-all-interfaces"
- add network:set subcommand
