Feature request: using recorded test data for more

[lcnbr]

Motivation

The recording feature is awesome, and replaces a janky local script that we had to record passing tests. However I couldn't find any tracking issue/ additional plans for it and I think there are many very useful additons that can be built on top of it.

Proposal

Here are a few ideas/ feature requests:

• Record the runnning time to be able to filter based on recorded timings (i.e. not kill slow tests, but just run the known fast ones)
• Being able to run already passing tests to test for regressions (if the current trunk has failing tests it is more useful to know that you haven't worsened the situation)

Alternatives

No response

Additional context

No response

[sunshowers]

Thanks for filing this issue! Glad recording has been working out for you; the intention of the feature is to create a platform on which much can be built.

Record the runnning time to be able to filter based on recorded timings (i.e. not kill slow tests, but just run the known fast ones)

Yeah this makes sense.

Being able to run already passing tests to test for regressions (if the current trunk has failing tests it is more useful to know that you haven't worsened the situation)

Interesting, I hadn't considered this before! In general I think I'm a green trunk kind of gal :)

[lcnbr]

Aha Yes I would love to be in green trunk world! Also maybe the filterset DSL could be state aware, so you could do smth like cargo nextest run -E 'recorded(runset)' where runset is some way to specify a runset like the jj revset DSL. And you could integrate additional flags, say if you have a recorded run of id 90d... then recorded(90d) is all tests run during the run, recorded(passing(90d)) is all passing tests, recorded(failing(90d)) all failing tests, recorded(faster-than(10s,90d)) is all tests that took less than 10s to run for the run 90d etc!