This is an attempt to write a log capture integration for the functional test runner that will collect all log output written during a test, prevent it from being written to stdout to lower the amount of work Jenkins master has to do and include the entire log output for each test in the junit report. I'm slightly concerned about the amount of memory we'll be using to store a copy of the logs for each test, but streaming the XML report isn’t easy and there doesn’t seem to be a library out there to help. If we decide it’s torally necessary we might be able to figure it out.
* [npm] upgrade to RxJS 6 (#18885)
This PR upgrades RxJS to version 6 and switches to a fork of `stream-to-observable` which includes an updated version of `any-observable` that supports RxJS 6 until https://github.com/jamestalmage/stream-to-observable/pull/10 is merged. The primary change in this version of RxJS is the movement of stream operators from `Obersable.prototype` to the `rxjs/operators` module. Some of the operators, like `catch` and `do`, have been renamed (`catchError`, and `tap`). The Obsevable factories have also been moved from static methods on the `Observable` class to named exports of the root `rxjs` module. Some of those factories have also changed slightly, like `fromEvent` which now emits arrays if the event handler is called with multiple arguments.
```js
// import the Rx namespace to get the Observable factories
import * as Rx from 'rxjs';
// import the operators as named imports
import { map, tap, switchMap } from 'rxjs/operators';
```
* [rxjs/dev-utils] fix old operator usage
* [rxjs/dev-utils] remove one more old operator
Restructure testing with kbn-test package
- Run with multiple configs, move cli options to config
- Package-ify kbn-test
- Eventually we'll have functional_test_runner live in a package
of its own, and then this kbn-test will use that as a dependency,
probably still as a devDependency.
- Implement functional_tests_server
- Collapse single and multiple config apis into one command
Use kbn-es
Replace es_test_cluster + es_test_config with kbn/test utils
Implement new createEsTestCluster
Improve scripts, jsdocs, cli top-level tools
Lift error handling to the top level
Restructure testing with kbn-test package
- Run with multiple configs, move cli options to config
- Package-ify kbn-test
- Eventually we'll have functional_test_runner live in a package
of its own, and then this kbn-test will use that as a dependency,
probably still as a devDependency.
- Implement functional_tests_server
- Collapse single and multiple config apis into one command
Use kbn-es
Replace es_test_cluster + es_test_config with kbn/test utils
Implement new createEsTestCluster
Improve scripts, jsdocs, cli top-level tools
Lift error handling to the top level
* [x-pack/ftr] call fatalErrorHandler when functional tests fail
* [kbn/dev-tools/withProcRunner] require a log as the first arg
* [kbn/dev-tools/procRunner] use correct promise, convert to getter
* [x-pack/ftr] avoid race condition that prevents success message logging
When starting the kibana server it is possible for log messages to come
after the server is started, so we added a pause that waits for 5
seconds of logging silence before logging the success message. The
observable used fails to complete though if a log message is never
written AFTER the Kibana server starts. To counter this the observable
is started with `null` so it will always start at least one 5 second
timer and always complete even if there is no log data after Kibana
server starts.
* fix typo
* [kbn-dev-utils/procRunner] try using execa to avoid never-exitting procs
* [kbn-dev-utils/procRunner] don't listen for close event, rely on stdio streams ending
In an attempt to debug random stalls we are seeing in CI, the proc runner will now send SIGKILL if a process does not exit after receiving the standard exit signal it receives (usually SIGTERM). Additionally, if the child process object never emits "close" or "exit" another timeout will trigger an error