[Development] qt5 modules in namespaces

[Olivier Goffart]

On Friday 30 December 2011 14:55:55 Denis Dzyubenko wrote:
> Hi guys,
> 
> With a qt5 module that we are developing, qtjsondb, we stumbled upon
> an issue with moc and our classes that are inside a namespace. Let me
> illustrate it with a code:
> 
> jsondb-global.h:
> 
> #if defined(QT_NAMESPACE)
> #  define QT_ADDON_JSONDB_BEGIN_NAMESPACE namespace QT_NAMESPACE {
> namespace QtAddOn { namespace JsonDb {
> #  define QT_ADDON_JSONDB_END_NAMESPACE } } }
> #  define QT_ADDON_JSONDB_USE_NAMESPACE using namespace
> QT_NAMESPACE::QtAddOn::JsonDb;
> #else
> #  define QT_ADDON_JSONDB_BEGIN_NAMESPACE namespace QtAddOn { namespace
> JsonDb { #  define QT_ADDON_JSONDB_END_NAMESPACE } }
> #  define QT_ADDON_JSONDB_USE_NAMESPACE using namespace QtAddOn::JsonDb;
> #endif
> 
> jsondb-map-reduce.h
> 
> QT_ADDON_JSONDB_BEGIN_NAMESPACE
> class Foo : public QObject {
> Q_OBJECT
> };
> QT_ADDON_JSONDB_END_NAMESPACE
> 
> 
> and here lies the problem - the macro QT_ADDON_JSONDB_BEGIN_NAMESPACE
> expands to a namespace definition, but moc doesn't know about that, so
> the generated moc_jsondb-map-reduce.cpp" expects to find class Foo in
> a global namespace:
> 
> moc_jsondb-map-reduce.cpp:
> 
> QT_BEGIN_MOC_NAMESPACE
> void Foo::qt_static_metacall(QObject *_o, QMetaObject::Call _c, int
> _id, void **_a)
> ...
> 
> 
> notice the QT_BEGIN_MOC_NAMESPACE - that is a trick for the case when
> Qt itself is compiled in a namespace - it is defined in qglobal.h to
> "using namespace QT_NAMESPACE"
> 
> In qt5 modules we have exactly the same problem as in namespaced-Qt
> itself. Hence my suggestion to solve it in the same way - I would like
> to be able to re-use that QT_BEGIN_MOC_NAMESPACE for qt5 modules, to
> be able to redefine it to use both Qt namespace and my modules'
> namespace.
> 
> Here is the patch for qtbase: http://codereview.qt-project.org/11892
> 
> And here is the patch for qtjsondb: http://codereview.qt-project.org/11894
> (see
> http://codereview.qt-project.org/#patch,sidebyside,11894,1,src/daemon/daemo
> n.pri for most interesting part :)

I wish moc had proper support for macro expension. so this would not be 
required at all.