Recover in suite

[YaroslavDev]

IMHO, panic in tests is not a good idea. Not only because it's hard to understand what test panicked, but panic also crashes entire test run and some tests don't even start. I think it's a good idea to isolate each test run by putting a defer/recover so that you know exactly what test caused panic and give possibility for other tests to run.

Github PR implementing this feature: #696

[iaroslav-ciupin]

P.S.
If you want to test that some of your code panics as expected, you must explicitly put a recover in your test, catch actual panic and compare it with expected one.
IMHO,
Panic SHOULD NOT propagate outside scope of a test case, because in a context of testing panic is just another type of error that should be caught and checked.

[itzamna314]

+1 to this. With test suites that have some environmental setup/teardown, a panic will exit tests before the teardown can happen. Then you've got resources you have to clean up manually.

Real-world example: I have a test suite that sets up a docker database container, then runs tests against it, and removes the container plus all volumes when everything is complete. If I have a panic in my code, I have to manually go through and clear out the container and its volumes.

[YaroslavDev]

Merged