Instead use sigc::mem_fun(reference, func).

This shows how having a sigc::notifiable base class is much nicer
than using these objects as void* as in libsigc++-2.0.

sigc::bind<1>() is meant t bind a value for the second argument,
which makes no sense for a method with 1 parameter.
sigc::bind() seems fine.

To deal with this clang warning when building with --enable-warnings=fatal.
TODO: However, this doesn't actually seem to disable the error.

glibmm_interface_move/main.cc:9:1: error: unused function 'glib_autoptr_cleanup_TestIface' [-Werror,-Wunused-function]
G_DECLARE_INTERFACE(TestIface, test_Iface, TEST, IFACE, GObject)
^
/opt/gnome/include/glib-2.0/gobject/gtype.h:1565:3: note: expanded from macro 'G_DECLARE_INTERFACE'
  _GLIB_DEFINE_AUTOPTR_CHAINUP (ModuleObjName, PrerequisiteName)                                           \
  ^
/opt/gnome/include/glib-2.0/glib/gmacros.h:421:22: note: expanded from macro '_GLIB_DEFINE_AUTOPTR_CHAINUP'
  static inline void _GLIB_AUTOPTR_FUNC_NAME(ModuleObjName) (ModuleObjName **_ptr) {                     \
                     ^
/opt/gnome/include/glib-2.0/glib/gmacros.h:415:43: note: expanded from macro '_GLIB_AUTOPTR_FUNC_NAME'
                                          ^
<scratch space>:224:1: not...

Putting the comment before the code, instead of to the right,
because clang-format can not (and could not) know which comments
are meant to be aligned. I prefer this anyway.

This seems to work.

I need to make a subsequent commit to fix the now-unaligned
trailing comments.

Because it sometimes wrongly things that comments on two lines
are related.

This now wrongly unaligns some blocks of trailing comments,
but I would rather just not have them.

Because this can break builds.

For automatic reformatting of the source code.

* tools/m4/vfunc.m4: Propagate a Glib::Error exception to GError** error.
Don't throw an exception in the callback function, called from C code.
Bug #762437

By changing
) dnl
to
)dnl

Because I noticed that this has just been documented as being allowed
to be null:
https://git.gnome.org/browse/glib/commit/?id=a4ed89bf75190631087596be22f9eb10dc6e53ef

This makes it easier to try C++14 sometimes and eventually really switch
to it.

* gio/src/dbusconnection.[hg|ccg]: Some string parameters in some C functions
accept nullptr, others don't.
call(), call_sync(): bus_name: Pass nullptr for an empty string.
emit_signal(): Only destination_bus_name can be nullptr.
               Other string parameters: Pass c_str().
The bug in call() and call_sync() was noticed by Gerlof Fokkema on gtkmm-list.
The bug in emit_signal() is an overreaction to bug 645072. All string
parameters were changed to pass nullptr, but only destination_bus_name
should have been changed.

* tools/pm/Function.pm:
* tools/pm/Output.pm: Accept {NULL} or {!NULL} after the name of a string
parameter in _WRAP_METHOD. It specifies how to translate an empty string to
a C string. {NULL}: nullptr, {!NULL}: pointer to an empty string.
Both {NULL} and {!NULL} can be useful, because the default translation is
different for mandatory and optional ({?}) parameters.

* glib/glibmm/dispatcher.[cc|h]: Add Dispatcher::connect(&&).
* glib/glibmm/objectbase.[cc|h]: Add ObjectBase::connect_property_changed(&&)
and ObjectBase::connect_property_changed_with_return(&&).
* glib/glibmm/propertyproxy_base.[cc|h]: Add PropertyProxyConnectionNode(&&)
ctor, PropertyProxyConnectionNode::connect_changed() and
SignalProxyProperty::connect(&&).
* glib/glibmm/signalproxy.[cc|h]: Add SignalProxyNormal::connect_impl_(&&),
SignalProxy::connect(&&), SignalProxy::connect_notify(&&),
SignalProxyDetailed::connect_impl_(&&), SignalProxyDetailedAnyType::connect(&&)
and SignalProxyDetailedAnyType::connect_notify(&&).
* glib/glibmm/signalproxy_connectionnode.[cc|h]:
Add SignalProxyConnectionNode(&&) ctor

Bug #761046