Repository navigation
Support force stop of running tests #1941
Description
Activity
This is because when you interrupt a test, it will finish running the currently pending tests and finish uploading the results. This is the clean way of performing a shutdown of a running test.
I suppose we could potentially also add support for forcing the shutdown of the test immediately (maybe the user has to tap again "stop the test now"), but in general I think we should be discouraging this.
- changed the title
[-]Make the interruption of a test immediate[/-][+]Support force stop of running tests[/+]on Jan 19, 2022 @hellais I think we should respect what the users want more aggressively. I know it's a shame interrupting a test midway, but it's also annoying to users if a test runs for a long time. Because of timeouts and submission, interrupting a running test could actually take a minute or more (if summing every possible bad-luck event together).
Ideally, tapping twice would be the right thing to implement. Given that doing that is not super simple to implement, I am actually going to suggest we refactor the code to religiously honor a context and bail as soon as a context is explicitly canceled. Because we also have background runs, I think it is not worth it wrestling with extra complexity to implement smooth test interruption.
Another middle ground implementation, instead, could be that, when the user taps interrupt we arm a five second timer and we say to the user "giving five seconds to the current test to finish". And then, after the give seconds, we cancel the context and cause the experiment to stop at the point at which it was running, regardless of which that point was.
Your thoughts?
@agrabeli I would like to know whether this issue is desktop, cli, mobile, or all of them. It would quite useful to know to assign it a scope.
This is now implemented as part of multiplatform changes.
During the OONI Partner Training, a participant said that the interruption of an OONI Probe test is not immediate, and that it takes some time to terminate a running test. Can we look into this?