testDelegateRegionDidChange is too fragile

[kkaefer]

I'm seeing relatively frequent test failures, but they're not reproducible. It looks like they heavily depend on timing.

See https://travis-ci.org/mapbox/mapbox-gl-native/jobs/60350680#L1051

@incanus @1ec5

[friedbunny]

Says it's doing two unanimated changes when it should only be performing one.

testDelegateRegionDidChange, ((unanimatedCount) equal to (1)) failed: ("2") is not equal to ("1") - regionDidChange delegate should indicate one unanimated change

test/ios/MapViewTests.m#L416

[friedbunny]

I think L416's regionDidChange delegate should indicate one unanimated change is getting bogged down and executing after the next test case has started.

If that's correct, then this should fix it:

diff --git a/test/ios/MapViewTests.m b/test/ios/MapViewTests.m
index fff52ad..08a867d 100644
--- a/test/ios/MapViewTests.m
+++ b/test/ios/MapViewTests.m
@@ -406,7 +406,7 @@
                                                    tester.mapView.centerCoordinate = CLLocationCoordinate2DMake(0, 0);
                                                }];

-    [tester waitForTimeInterval:1];
+    [tester waitForAnimationsToFinishWithTimeout:1];

     XCTAssertEqual([notification.userInfo[@"animated"] boolValue],
                    NO,
@@ -415,6 +415,8 @@
                    1,
                    @"regionDidChange delegate should indicate one unanimated change");

+    [tester waitForTimeInterval:1];
+
     notification = [system waitForNotificationName:@"regionDidChangeAnimated"
                                             object:tester.mapView
                                whileExecutingBlock:^{

[friedbunny]

Noooope, other way around. The test is failing because it's not waiting long enough for the environment to be reset and it's catching that.

[incanus]

This is looking great when testing locally and we can probably roll this into reenabling CI iOS tests in #1452. We'll give everything some care then to avoid false-positives like this one.