Goal:

Know if there's something that was left on the GUI context for application testing. A good example is the Windows Error Report dialog (WER).


AIs:

When the following conditions are true:
- running in the context of a GUI
- upon task failure, in particular in the context of zombie processes

- task_runner takes a main display screen shot and saves it as task metadata via "amend", likely as TaskOutput(id=screen.png)


Alternative:

- Use a common wrapper script for all the Chromium tasks to do it instead and save the screenshot in the isolated directory. It's generally better to do it this way but the main downside is that it is not possible to catch issues with zombie process detection.