I haven't tested it at local yet. But from the screen record, it seems that more features are needed to make it more useful than a shell script. See #162 (comment).

• A button to add the counterexample in the test cases.
• Be able to choose gen and std from editor tabs (as well as local files).
• Be able to use a code template dedicated to generators.

It is almost the same as Polygon's, so only testilb style generators are supported.

By "testlib style", do you mean generators that take command-line arguments? I don't think we need to force this style. Now we provide the Polygon functionality to feed random arguments to the generator, and testlib-style generators can make use of these arguments. But the user can keep the arguments empty and generate different test cases with the same empty arguments. And if we don't restrict the style, we can also support Python/Java generators.

I'm thinking that the generator argument design might be unintuitive for users unfamiliar with the testlib-style generator / generator with arguments. Maybe we can have two modes, one without arguments and one with an arguments pattern. In the no argument mode, I'm not sure whether it's better to set the number of test cases, to set the total running duration, or to run indefinitely until manual stop 🤔

This can be simulated by [1..n], but the user might not know this without hints. And maybe it's just simpler to run until manual stop. I think that's what I always did with a simple bash script.

I think in the no argument mode, we could let it run until it’s manually stopped, and display both the number of test cases that have been run and the elapsed time？

Generator and std chosen from tabs should be automatically saved depending on SettingsHelper::isSaveFileOnCompilation().

Thanks for sticking around and driving this to completion.