Robots: Continuous Integration Support


test-patch works hand-in-hand with various CI and other automated build systems. test-patch will attempt to auto-determine if it is running under such a system and change its defaults to match known configuration parameters automatically. When robots are activated, there is generally some additional/changed behavior:

Automated Robot Detection

The following systems are detected automatically when run in one of these environments:

Manual Configuration

For automated systems that are not directly supported, --robot tells test-patch that this is an automated system. This will trigger many of the above settings.

The --build-url option is also useful when running in --robot mode so that emails and such
have a location to look at the output artifacts:

$ test-patch --robot --build-url=${buildnumber}/

Some plug-ins such as Maven have special handling if there are multiple executions of test-patch happening at once. It is very common when using automation systems to have multiple runs on the same host. In order to assist these plug-ins, an instance identifier may be provided:

$ test-patch --robot --instance=1

If --robot is specified without an instance, a random number is generated and used.

Sentinel Mode

If stuck Docker containers are a problem, a more aggressive robot may be enabled with the --sentinel option. This option
enables killing containers that have been running for over 24 hours as well. See the documentation on docker
for more information.

Continuous Improvement

It isn't unusual when adding a new linter to an existing source repository to be overwhelmed with new errors.
Most projects want to start out where all new patches cannot add new errors, but the existing ones are
acceptable. In order to enable branches to build with errors but patches/PRs to fail on robots,
test-patch supports the --continuous-improvement flag:

Normally, test-patch execute with this flag set to false:

$ test-patch --continuous-improvement=false (other options)

This setting will fail branches and PRs if any errors are found. However, if this value is set to true:

$ test-patch --continuous-improvement=true (other options)

then branches will still report errors but the return code will be successful. Using this mode is ideal for
many CI/CD systems that are using repositories with known problems.