Checks in some tests are broken

[vaind]

I've noticed while working on recent PRs that test code sometimes uses expect() and other types of assertions in callbacks to Sentry. E.g. Sentry.init() - these are broken and are actually not being checked at all because Sentry swallows these errors (doesn't propagate them). Minimal repro - the following test passes:

  test('should fail', () async {
    await Sentry.init((options) {
      fail('hello there');
    }, options: SentryOptions(dsn: fakeDsn));
  });

Besides a testing issue, this may also be a design problem - naively, as a User, I'd expect exceptions in user-provided code to be propagated.

[ueman]

Related #990

[marandaneto]

@vaind Sentry should swallow every exception when a user makes a mistake in the callbacks, that's intended and by design.
We'd need to fix those tests for sure, mind checking this out? or giving an example (a test's name)?

[marandaneto]

Related #990

Likely an oversight because of PRs related to this, indeed.

[marandaneto]

@vaind Sentry should swallow every exception when a user makes a mistake in the callbacks, that's intended and by design. We'd need to fix those tests for sure, mind checking this out? or giving an example (a test's name)?

@vaind in case you remember where you have seen this.

[vaind]

e.g. multiple calls to await Sentry.init( in sentry_test.dart have test assertions inside the callback

[vaind]

I'd suggest having a test-mode/dev-mode setting where exceptions are not captured in callbacks. This would could also help users if they're having trouble with reporting some errors, as it's much easier to find errors this way than look for logs.