Qt-interest Archive, April 2001
WG: widget-destructor
Message 1 in thread
To trolls and qt-users:
in QWidget::~QWidget() why
if ( childObjects ) { // delete children objects
QObject* obj;
QObjectListIt it(*childObjects);
QObject *obj;
while ( (obj=it.current()) ) {
++it;
obj->parentObj = 0;
// ### nest line is a QGList workaround - remove in 3.0
childObjects->removeRef( obj );
delete obj;
}
and not
if ( childObjects ) { // delete children objects
QObject* obj;
while( obj= childObjects->first() ) delete obj;
}
???
That would solve the
parentless-non-toplevel-widget-during-destruction-problem because that's the
origin of the problem.
bye
Daniel
-----Ursprungliche Nachricht-----
Von: porten@trolltech.com [mailto:porten@trolltech.com]Im Auftrag von
Harri Porten
Gesendet: Mittwoch, 18. April 2001 16:34
An: rolf@finecode.de
Betreff: Re: widget-destructor
> Daniel Rolf wrote:
>
> Hello,
>
> I've got a problem:
> the intermediate state if a widget is being destroyed is fatal for me:
>
> if a widget is destroyed it destroys all its child-widgets by setting
> their parent to 0 and the deleting them. This is fatal because setting
> the parent to 0 will set the widget to a contradictive state because
> it's not a top-level-widget.
> qt calls my destructor and I call some internal code which results to
> some gp if calling setEnabled(FALSE) because the widget is not
> toplevel neither it has a parent (anymore):
>
> void QWidget::setEnabled( bool enable )
> {...
> if ( !isTopLevel() && !parentWidget()->isEnabled() && enable
> )
> <----- here it fails because isTopLevel returns NO
> and parentWidget returns NULL
> ...
> }
>
> any idea? that should be fixed in the widget-destructor!!!!!!!!
Yes.Please try the attached patch.
But why do you call setEnabled() inside of a widget being destructed ?
:)
Harri.
==== //depot/qt/2.3/src/kernel/qwidget.cpp#1 (text) -
//depot/qt/2.3/src/kernel/qwidget.cpp#2 (text) ==== content
@@ -1266,7 +1266,8 @@
else
setWState( WState_ForceDisabled );
- if ( !isTopLevel() && !parentWidget()->isEnabled() && enable )
+ if ( !isTopLevel() && parentWidget() &&
+ !parentWidget()->isEnabled() && enable )
return; // nothing we can do
if ( enable ) {
Message 2 in thread
Daniel Rolf wrote:
>
> To trolls and qt-users:
>
> in QWidget::~QWidget() why
>
> if ( childObjects ) { // delete children objects
> QObject* obj;
> QObjectListIt it(*childObjects);
> QObject *obj;
> while ( (obj=it.current()) ) {
> ++it;
> obj->parentObj = 0;
> // ### nest line is a QGList workaround - remove in 3.0
> childObjects->removeRef( obj );
> delete obj;
> }
>
> and not
>
> if ( childObjects ) { // delete children objects
> QObject* obj;
> while( obj= childObjects->first() ) delete obj;
> }
>
> ???
Setting the parent object of a to be destructed widget to 0 makes a lot
of sense. You neither want the child to call any functions of the parent
which is about to destruct itself nor send it any events.
> That would solve the
> parentless-non-toplevel-widget-during-destruction-problem because that's the
> origin of the problem.
Did the patch I send to you work, i.e. prevent the crash ?
Harri.
Message 3 in thread
Harri,
the patch works. But there are some more places where to attach the patch.
bye
Daniel
-----Ursprungliche Nachricht-----
Von: owner-qt-interest@trolltech.com
[mailto:owner-qt-interest@trolltech.com]Im Auftrag von Harri Porten
Gesendet: Freitag, 20. April 2001 03:52
An: rolf@finecode.de
Cc: qt-interest@trolltech.com
Betreff: Re: WG: widget-destructor
Daniel Rolf wrote:
>
> To trolls and qt-users:
>
> in QWidget::~QWidget() why
>
> if ( childObjects ) { // delete children objects
> QObject* obj;
> QObjectListIt it(*childObjects);
> QObject *obj;
> while ( (obj=it.current()) ) {
> ++it;
> obj->parentObj = 0;
> // ### nest line is a QGList workaround - remove in 3.0
> childObjects->removeRef( obj );
> delete obj;
> }
>
> and not
>
> if ( childObjects ) { // delete children objects
> QObject* obj;
> while( obj= childObjects->first() ) delete obj;
> }
>
> ???
Setting the parent object of a to be destructed widget to 0 makes a lot
of sense. You neither want the child to call any functions of the parent
which is about to destruct itself nor send it any events.
> That would solve the
> parentless-non-toplevel-widget-during-destruction-problem because that's
the
> origin of the problem.
Did the patch I send to you work, i.e. prevent the crash ?
Harri.
--
[ signature omitted ]
