Surprising behaviour of tests on Windows in some specific contexts about easystats HOT 5 CLOSED

@easystats/maintainers Any ideas?

from easystats.

I can't reproduce the failure for tests in random order but I suspect that the definition of test failure and test error doesn't work when there's only 1 test and that this test is skipped (I don't know why, it's the only thing I can think of). Btw, what's the difference between failure and error for testthat?

Removing the reporter = "silent" argument would probably give more info on this

from easystats.

Btw, what's the difference between failure and error for testthat?

• Failure: Actual value doesn't match with expected value
• Error: Test code itself couldn't be run successfully

So the random test order workflow is now passing. It was failing because Pandoc wasn't installed for this workflow, and flaxdashboard seems to require it.

I still don't understand why code coverage workflow is stuck in infinity.

from easystats.

I think covr doesn't take into account the skip_ calls. Locally, I tried to put all kind of skip_ in the test file. devtools::test() correctly skips all of them, but covr::package_coverage() stil returns the same percentage as if there were no skips. So maybe it stays stuck forever in one of the interactive tests?

from easystats.

Thanks! That was indeed the issue. Skipping these tests during code coverage fixes that workflow. I will investigate which is the offending test and why it gets stuck.

from easystats.