Add advanced browser example testplan Goal of this plan is to show how you can define a testplan which can be run in a node environment as well as within a browser, without having to change anything to your testplan code.

This does mean we also need to allow the @testground/sdk javascript package to be run in the browser as-is. To allow this I have a branch which I run this testplan against at https://github.com/glendc/sdk-js/tree/feat/support-browser:

• the FS features are not possible in the browser, so these need to be disabled;
• the environment variables are not available when creating the runtime from env in browser,
  so for the browser we instead assume that these are available in the window object as window.testground.env;

For both these changes we assume one is in the browser when process.title === 'browser', this because the browser npm (polyfill webpack) module sets that as the process global variable.

Within the example the testplan is found under src/, the other three directories form together the "magic" solution to allow the node testplan to also be run in the browser:

• scripts/: contains a polyfill folder to inject an import in src/index.js which is required to polyfill some logic required by winston (an import by @testground/sdk)
• browser/: the entry point for the docker:generic testplan, and which will spawn and control everything else;
  • browser/server/: simple express served — controlled by the browser/ and which will serve the web-pack'd testplan;

TODO:

• allow debug breakpoints in chrome://inspect without having to refresh; (not an issue for me, and don't think there is a way to do it otherwise)
• make the tests actually run and pass (as for now we no longer have errors and everything seems polyfill'ed, but the test doesn't seem to run (despite the fact that we do see logs coming through);
• open a @testground/sdk MR and align on the approach there: my goal is that this sdk can be used just the same from the browser as well as the node env;
• make the browser magic solution more smart in terms of how to wait for test to finish: somehow hook into an event which will be called when testplan is finished, not before and not later;

Related PRs:

• add browser support to @testground/sdk: first baby steps towards native browser support for the JS sdk of testground sdk-js#25
  • replaced with new approach in @testground/sdk: feat: add basic browser support  sdk-js#26
• support docker runner (additional) environment arguments (required to allow the use of environment variables to choose what browser engine to test one, instead of having to rebuild the docker container): feat: support environment variables for local docker #1507

Potential improvements but probably out of scope:

• it would be great if we can choose what exposed port we bind to on the host machine (instead of currently any available open port) as it would mean we can always inspect against the same chrome address (but given this is defined by the docker Go library I do not have much hope for this one);
  • this one can be tackled in future, for now it is tracked in the following issue: host port choice for local:docker exposed ports #1506

In function of #1386.