This change allows operators to specify a DOCKER_BIN environment variable. This will specify a binary to run when executing docker, which is useful in cases where the 'docker' command being run must be modified in a way that would otherwise be invasive to Dokku, but minimalistic if done within a wrapper.
A DOKKU_SCALE file will always be used for managing the scale count for an app if it exists in a repository. As such, ps:scale usage should be disabled in those cases. This commit makes that more apparent.
Also document how to build a specific core plugin for testing purposes.
Closes#3544
If the subdomain is url-like, assume it is a valid url and use it for the default domain name. This allows users to specify the default domain for an app even if that is a subdomain of a global vhost.
Closes#3529
This change adds json output as both key/value as well as a list of objects. The former can be used in quick scripting environments, while the latter allows higher-level languages to have a bit more structure around how environment variables are declared. Specifically, systems such as kubernetes understand the latter method, while the former can be used within Nomad job files.
With the introduciton of procfile-util, scaling a non-existent process to 0 was disabled. This brings back that functionality, allowing users to scale all processes to 0. This is especially useful when a user has accidentally scaled a non-existent process up, causing deploys to fail without any way to revert outside of manually editing files.
Closes#3428
The `domains` command previously had this functionality, but is deprecated, and thus we should provide an alternative method of presenting the information.
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 .
```
The checkstyle xslt wasn't quite good enough, resulting in missing/malformatted output. Note that this now excludes _passing_ tests, which also isn't quite right, but a fair bit closer to the xsd here: https://github.com/windyroad/JUnit-Schema
This requires us to run shellcheck twice, as the first pass doesn't output anything human readable - both because of the redirect and because shellcheck doesn't directly support it - which increases lint runtime but allows us to have both the junit information and the normal human-readable output.
Note that the checkstyle output does not contain timing info.
