Summary: | blank diagram is erroneously marked modified | ||
---|---|---|---|
Product: | [Applications] umbrello | Reporter: | Oliver Kellogg <okellogg> |
Component: | general | Assignee: | Umbrello Development Group <umbrello-devel> |
Status: | RESOLVED FIXED | ||
Severity: | normal | ||
Priority: | NOR | ||
Version: | unspecified | ||
Target Milestone: | --- | ||
Platform: | unspecified | ||
OS: | Linux | ||
Latest Commit: | Version Fixed In: | ||
Sentry Crash Report: | |||
Attachments: |
set modified patch.
same patch but without the color codes |
Description
Oliver Kellogg
2003-12-26 20:13:58 UTC
This is caused by the UMLAssociation destructor (5 in backtrace) calling setModified. I've no idea what deletes the association. I suspected 'delete pObject' in UMLDoc::loadUMLObjectsFromXMI() but it doesn't seem to be that. 0: /usr/lib/libkdecore.so.4(_Z11kdBacktracei+0x45) [0x4078bb85] 1: /usr/lib/libkdecore.so.4(_Z11kdBacktracev+0x2b) [0x4078bf0b] 2: umbrello(_ZN6UMLApp11setModifiedEb+0x9e) [0x815026e] 3: umbrello(_ZN6UMLDoc11setModifiedEbb+0x45) [0x8161db5] 4: umbrello(_ZN6UMLDoc17removeAssociationEP14UMLAssociation+0x49) [0x815a979] 5: umbrello(_ZN14UMLAssociationD0Ev+0x4f) [0x80dd7cf] 6: /usr/lib/libqt-mt.so.3(_ZN7QObject5eventEP6QEvent+0xac) [0x40bc36fc] 7: /usr/lib/libqt-mt.so.3(_ZN12QApplication14internalNotifyEP7QObjectP6QEvent+0xbf) [0x40b6e57f] 8: /usr/lib/libqt-mt.so.3(_ZN12QApplication6notifyEP7QObjectP6QEvent+0x10b) [0x40b6db8b] 9: /usr/lib/libkdecore.so.4(_ZN12KApplication6notifyEP7QObjectP6QEvent+0xaf) [0x4076ff3f] 10: /usr/lib/libqt-mt.so.3(_ZN12QApplication16sendPostedEventsEP7QObjecti+0x176) [0x40b6f346] 11: /usr/lib/libqt-mt.so.3(_ZN12QApplication16sendPostedEventsEv+0x28) [0x40b6f1c8] 12: /usr/lib/libqt-mt.so.3(_ZN10QEventLoop13processEventsEj+0x7e) [0x40b1dbfe] 13: /usr/lib/libqt-mt.so.3(_ZN10QEventLoop9enterLoopEv+0xdf) [0x40b8039f] 14: /usr/lib/libqt-mt.so.3(_ZN10QEventLoop4execEv+0x24) [0x40b80254] 15: /usr/lib/libqt-mt.so.3(_ZN12QApplication4execEv+0x21) [0x40b6e7c1] 16: umbrello(main+0x18e) [0x8133e4e] 17: /lib/libc.so.6(__libc_start_main+0xc6) [0x41240da6] 18: umbrello(_ZN6QFrame10paintEventEP11QPaintEvent+0x31) [0x80d91c1] ] Subject: Re: [Uml-devel] New: blank diagram is erroneously marked modified Oliver Kellogg <okellogg@users.sourceforge.net> [031228 17:33]: > http://bugs.kde.org/show_bug.cgi?id=71265 > Summary: blank diagram is erroneously marked modified > ... > Start umbrello and open a diagram file. > Close the diagram without any modifications. > Umbrello will still ask "The diagram has been modified. Do > you wish to save..." This is especially annoying on the blank (default) diagram. I think the problem is fixed, because the setModified() call is already commented out. Nevertheless, please check again. Steinchen Still happens with cvs of today, 2004/05/11. Created attachment 5955 [details]
set modified patch.
contentsMouseRelease does a setModified. This should be done at the
toolbarstate-classes. Probably this bug reappeared after the
toolbarstate-patch.
Created attachment 5959 [details]
same patch but without the color codes
Sorry, the previous patch contains color-codes generated by a color_cvs script.
Here is the same patch again, but without these markup codes.
Raymond Bosman <bosman@hetnet.nl> [040512 08:36]: > contentsMouseRelease does a setModified. This should be done at the > toolbarstate-classes. I think the whole setModified() and Undo/Redo architectur is wrong. At the moment there are setModified() calls in data classes and in view classes. In my opinion such calls should only happen in view classes. What do you think the right solution would be? Sebastian The setModified() has nothing to do with the internal data (classes). So I agree that it should only happen in the view classes. Speaking about Model-View-Controller: Are all data-classes 'view-code' free? (For example the class: AssociationWidget) Thanks for your patch, it's in CVS. See question 3 at http://uml.sourceforge.net/faq.php for the best way to make a patch (i.e. "cvs diff"). Umbrello's classes are in many parts suffering from bitrot, the model-view-controller idea is probably not followed in various places (all more patches happily received :) |