Initializes an environment for debugging the browser tests. Includes an dedicated instance of the kibana server for building the test bundle, and a karma server. When running this task the build is optimized for the first time and then a karma-owned instance of the browser is opened. Click the "debug" button to open a new tab that executes the unit tests.
Run single tests by appending `grep` parameter to the end of the URL. For example `http://localhost:9876/debug.html?grep=ML%20-%20Explorer%20Controller` will only run tests with 'ML - Explorer Controller' in the describe block.
The functional UI tests, the API integration tests, and the SAML API integration tests are all run against a live browser, Kibana, and Elasticsearch install. Each set of tests is specified with a unique config that describes how to start the Elasticsearch server, the Kibana server, and what tests to run against them. The sets of tests that exist today are *functional UI tests* ([specified by this config](test/functional/config.js)), *API integration tests* ([specified by this config](test/api_integration/config.js)), and *SAML API integration tests* ([specified by this config](test/saml_api_integration/config.js)).
The script runs all sets of tests sequentially like so:
* builds Elasticsearch and X-Pack
* runs Elasticsearch with X-Pack
* starts up the Kibana server with X-Pack
* runs the functional UI tests against those servers
* tears down the servers
* repeats the same process for the API and SAML API integration test configs.
API integration tests are intended to test _only programmatic API exposed by Kibana_. There is no need to run browser and simulate user actions, which significantly reduces execution time. In addition, the configuration for API integration tests typically sets `optimize.enabled=false` for Kibana because UI assets are usually not needed for these tests.
We also have SAML API integration tests which set up Elasticsearch and Kibana with SAML support. Run API integration tests separately with SAML support like so:
If you are **developing functional tests** then you probably don't want to rebuild Elasticsearch and wait for all that setup on every test run, so instead use this command to build and start just the Elasticsearch and Kibana servers:
For both of the above commands, it's crucial that you pass in `--config` to specify the same config file to both commands. This makes sure that the right tests will run against the right servers. Typically a set of tests and server configuration go together.
Read more about how the scripts work [here](scripts/README.md).
For a deeper dive, read more about the way functional tests and servers work [here](packages/kbn-test/README.md).
You may see an error like this when you are getting started:
```
[14:08:15] Error: Linux x86 checksum failed
at download_phantom.js:42:15
at process._tickDomainCallback (node.js:407:9)
```
That's thanks to the binary Phantom downloads that have to happen, and Bitbucket being annoying with throttling and redirecting or... something. The real issue eludes me, but you have 2 options to resolve it.
1. Just keep re-running the command until it passes. Eventually the downloads will work, and since they are cached, it won't ever be an issue again.
1. Download them by hand [from Bitbucket](https://bitbucket.org/ariya/phantomjs/downloads) and copy them into the `.phantom` path. We're currently using 1.9.8, and you'll need the Window, Mac, and Linux builds.