See the qtbug30200.log . Note:

• It seems to succeed in the CI (apparently with GLIB-eventloop?)
• The test is borked in that it creates another object whose code contains QVERIFY(). That is, the test slot does not quit if a failure occurs (and later hangs).

The diagnostic patch qtbug30200_diag.patch adds the same test to tst_qeventloop, where it succeeds (with a QCoreApplication); the destructor of the EventLoopNester instance is called after the check.

It also removes the other tests from the slot testDeleteLaterProcessEvents of tst_qapplication to make things clearer. With the patch applied, it appears the ~EventLoopNester is called from qDeleteInEventHandler() before the pointer check (see qtbug30200_eventloopnesterdestructorstack.txt ).

Oddly enough, the test succeeds on Windows.