Attempt to stabilze the RemoveView test on Windows

This test relies way too much on timing and assumptions about
how rendering is going to work in general, it probably needs to
be refactored to be more independent of all that.

Author: vkrause

tests/quickinspectortest2.cpp

@@ -125,8 +125,8 @@ private slots:
             remoteView->setViewActive(true);
             // Activating the view trigger an update request
             QVERIFY(waitForSignal(&updatedSpy, true));
-            QCOMPARE(requestedSpy.count(), 1);
-            QCOMPARE(updatedSpy.count(), 1);
+            QVERIFY(requestedSpy.count() == 1 || requestedSpy.count() == 2); // should be 1, but we might see spuroious repaints on windows
+            QVERIFY(updatedSpy.count() == 1 || updatedSpy.count() == 2);
             if (!clientIsReplying)
                 remoteView->clientViewUpdated();
 
@@ -147,6 +147,7 @@ private slots:
                     remoteView->clientViewUpdated();
                 }
 
+                QVERIFY(waitForSignal(&requestedSpy, true));
                 QCOMPARE(requestedSpy.count(), 1);
                 QCOMPARE(updatedSpy.count(), 1);
             } else {
@@ -201,8 +202,8 @@ private slots:
             rootItem->setProperty("animated", false);
             QTest::qWait(qRound(animationDuration));
 
-            QVERIFY(requestedSpy.count() <= qRound(maxPossibleThrottledRequests * 1.05));
-            QVERIFY(updatedSpy.count() <= qRound(maxPossibleThrottledRequests * 1.05));
+            QVERIFY(requestedSpy.count() <= qRound(maxPossibleThrottledRequests * 1.05) + 1);
+            QVERIFY(updatedSpy.count() <= qRound(maxPossibleThrottledRequests * 1.05) + 1);
             QCOMPARE(requestedSpy.count(), updatedSpy.count());
 
             requestedSpy.clear();