qApp: Ensure QtCore import when embedded QApplication subclass is used

The qApp machinery works great with Python.
When using embedding, things are different because there
is no longer a wrapper layer.

Unfortunately, many extension modules use C++ to derive a
QApplication class.

This has the side effect that when a foreign C++ module gets
imported, the qApp machinery does not see it as it would
in Python.

Instead of a complex analysis, we always make sure that QtCore
is imported. It will report the right instance, anyway.

This change could not easily be tested.
It was confirmed as a solution by Antonio Rojas.

Change-Id: Ie9c56ac75e6c0ae3ace615dfc26c6d218ff4efea
Fixes: PYSIDE-1135
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@qt.io>

1 files changed, 8 insertions, 1 deletions

diff --git a/sources/shiboken2/libshiboken/qapp_macro.cpp b/sources/shiboken2/libshiboken/qapp_macro.cpp
index 306f53b74..c2018bd6f 100644
--- a/sources/shiboken2/libshiboken/qapp_macro.cpp
+++ b/sources/shiboken2/libshiboken/qapp_macro.cpp
@@ -246,7 +246,14 @@ NotifyModuleForQApp(PyObject *module, void *qApp)
      * qApp_contents variable and assigns the instance, instead of vice-versa.
      */
     PyObject *coreDict = qApp_moduledicts[1];
-    if (qApp != nullptr && coreDict != nullptr) {
+    if (coreDict == nullptr) {
+        // PYSIDE-1135: Make sure that at least QtCore gets imported.
+        // That problem exists when a derived instance is created in C++.
+        qApp_moduledicts[1] = Py_None; // anything != nullptr during import
+        coreDict = PyImport_ImportModule("PySide2.QtCore");
+        qApp_moduledicts[1] = coreDict;
+    }
+    if (qApp != nullptr && coreDict != nullptr && coreDict != Py_None) {
         PyObject *coreApp = PyDict_GetItemString(coreDict, "QCoreApplication");
         if (coreApp != nullptr) {
             qApp_content = PyObject_CallMethod(coreApp, "instance", "");