[bt-devel] SegFault in 2.3.3 during navigation

[Martin Gruner]

Gary,

I am not sure about this. There are instances where it makes sense that 
different objects share the same key. That is because the key also holds a 
reference to a particular module, which is important for versification schemes 
and localization of bible booknames. At least that's the idea behind the 
current implementation, we might come up with something better, of course. One 
display window has one main key which is also used by the keychooser and its 
widgets etc. I keep forgetting the implementation details. :-(

mg

Am Mittwoch, 4. November 2009 21:21:24 schrieb Gary Holmlund:
> I looked at this for some time. I believe it is safe to copy the pointer
> for now.
> 
> The real problem here is where multiple classes use the same pointer to
> a verse reference. For instance, the CKeyReferenceWidget which
> implements the spinboxes for book, chapter, and verse. It keeps a copy
> of this pointer, increments it when the user changes the widget, and
> emits a signal for a verse change. The signal for verse change is caught
> by CBibleKeyChooser where it emits a beforeKeyChange using the same
> shared pointer. It does not know that the value in the pointer has
> already been changed. It then saves the pointer and emits a keyChanged
> with the saved pointer thinking that the second emit has a different
> value. This lead to problems in the saving of text in personal
> commentaries. Also, keeping a pointer as a member of multiple classes
> clearly can cause problems if any class decides to delete and recreate
> the pointer.
> 
> I think the real answer is to eliminate the use of a shared pointer for
> verse references. Each class should have a method of getting the verse
> reference value gets a non-pointer copy of the verse reference. All of
> the emit signals should pass a const reference to a verse value rather
> than the shared pointer. Doing this may take some time and perhaps
> should not be done as we are getting ready to make a release.
> 
> Gary
> 
> Martin Gruner wrote:
> > Eeli,
> >
> > I just looked at the "fix" in r1726. It seems kindof dangerous to just
> > copy the key object, introducing a memory leak. I would guess that this
> > only masks the original crash problem instead of fixing it, but I cannot
> > tell for sure now.
> >
> > I don't have the time now to investigate it, will be away until Monday.
> >
> > mg
> >
> > Am Mittwoch, 4. November 2009 17:10:12 schrieb Eeli Kaikkonen:
> >> On Wed, 4 Nov 2009, John A. Sullivan III wrote:
> >>> Hello, all.  It's great to be able to use BibleTime once again.
> >>> However, I am experiencing a readily reproducible segfault after
> >>> upgrading to 2.3.3 via an apt-get update on fully patched Ubuntu 8.04
> >>> on 64-bit AMD.  If I scroll forward a chapter using the next chapter
> >>> navigation arrows in my Bible pane and then click on a verse lower down
> >>> in the text, I get a segmentation fault every time.
> >>
> >> This has accidentally been fixed with svn r1726 while fixing another
> >> crash. It should be backported to 2.3 series.
> >>
> >>> I also noticed this wiped out any bookmarks I had created.  I looked
> >>> for a way to save the bookmarks other than quitting BibleTime as a way
> >>> to protect my work from this bug.  Perhaps I didn't look hard enough
> >>> but I did not see a way. Is there a way to save bookmarks from within a
> >>> session? Should they be save automatically upon creation so they are
> >>> not lost in the event of a crash? Thanks - John
> >>
> >> That's a reasonable request (saving automatically), but could you add it
> >> to our feature request database in sourceforge.net?
> >>
> >>   Yours,
> >> 	Eeli Kaikkonen (Mr.), Oulu, Finland
> >> 	e-mail: eekaikko at mailx.studentx.oulux.fix (with no x)
> >>
> >> _______________________________________________
> >> bt-devel mailing list
> >> bt-devel at crosswire.org
> >> http://www.crosswire.org/mailman/listinfo/bt-devel
> 
> _______________________________________________
> bt-devel mailing list
> bt-devel at crosswire.org
> http://www.crosswire.org/mailman/listinfo/bt-devel
> 

-- 

Bauplan des Lebens - längst im Gen entdeckt!
Die Wissenschaft ist stolz: Sie hat's gecheckt.
Nun ist der Bauplan als Beweis beliebt,
dass es den Architekten gar nicht gibt.

Wolf Rahn