Trolltech Home | Qt-interest Home | Recent Threads | All Threads | Author | Date
All threads index page 3

Qt-interest Archive, March 2008
Segmentation faults caused by QMap

Message 1 in thread

Hello everybody,

I have a problem with the the QMap class, following errors are coming:

Program received signal SIGSEGV, Segmentation fault.
0x000000000041da14 in QMap<int, TagItem*>::isEmpty (this=0x706f746b73654483) 
at /usr/include/qt4/QtCore/qmap.h:166
166         inline bool isEmpty() const { return d->size == 0; }

I know that at the point when isEmpty is called, the QMap is untouched by the 
program. Therefore it should return true but it just crashes. And I am not 
sure why.

Here the necessary code (reduced to the relevant parts):

void RDLDMap::addTagItem (int tagId, QColor color)
{
	if (tags.isEmpty() || tags.contains(tagId))
		return ;
...
}

class TagItem : public QObject, public QGraphicsEllipseItem
{
	Q_OBJECT
	public:
		TagItem ();
		void	setTagId (int id);
		int	getTagId ();

	signals:
		void mouseEnterTag (int tagId, int x, int y);
		void mouseLeaveTag ();
		void mouseMoveOverTag (int tagId, int x, int y);
		
	protected:
		void hoverEnterEvent (QGraphicsSceneHoverEvent * event);
		void hoverLeaveEvent (QGraphicsSceneHoverEvent * event);
		void hoverMoveEvent (QGraphicsSceneHoverEvent * event);

	private:
		int tagId;
};

class RDLDMap : public QGraphicsScene
{
	Q_OBJECT
...
	public slots:
		void addTagItem (int tagId, QColor color);
...
	private:
		QMap <int, TagItem *>	tags;
...
};

I tried it also without the isEmpty and only the contains, resulting into:
if (tags.contains(tagId))

Program received signal SIGSEGV, Segmentation fault.
0x000000000041d606 in QMap<int, TagItem*>::findNode (this=0x706f746b73654483, 
akey=@0x7fffafdac2c4) at /usr/include/qt4/QtCore/qmap.h:424
424         QMapData::Node *cur = e;

and i tried the direct access:
if (tags[tagId])

Program received signal SIGSEGV, Segmentation fault.
0x000000000041da58 in QMap<int, TagItem*>::detach (this=0x706f746b73654483) 
at /usr/include/qt4/QtCore/qmap.h:168
168         inline void detach() { if (d->ref != 1) detach_helper(); }

Thank you for your help in advance.

--
 [ signature omitted ]
Message 2 in thread

Your RDLDMap is probably null or previously deleted.



> -----Original Message-----
> From: SirTwist@xxxxxx [mailto:SirTwist@xxxxxx]
> Sent: Saturday, March 15, 2008 7:36 PM
> To: qt-interest@xxxxxxxxxxxxx
> Subject: Segmentation faults caused by QMap
> 
> Hello everybody,
> 
> I have a problem with the the QMap class, following errors are coming:
> 
> Program received signal SIGSEGV, Segmentation fault.
> 0x000000000041da14 in QMap<int, TagItem*>::isEmpty
> (this=0x706f746b73654483)
> at /usr/include/qt4/QtCore/qmap.h:166
> 166         inline bool isEmpty() const { return d->size == 0; }
> 
> I know that at the point when isEmpty is called, the QMap is untouched
> by the
> program. Therefore it should return true but it just crashes. And I am
> not
> sure why.
> 
> Here the necessary code (reduced to the relevant parts):
> 
> void RDLDMap::addTagItem (int tagId, QColor color)
> {
> 	if (tags.isEmpty() || tags.contains(tagId))
> 		return ;
> ...
> }
> 
> class TagItem : public QObject, public QGraphicsEllipseItem
> {
> 	Q_OBJECT
> 	public:
> 		TagItem ();
> 		void	setTagId (int id);
> 		int	getTagId ();
> 
> 	signals:
> 		void mouseEnterTag (int tagId, int x, int y);
> 		void mouseLeaveTag ();
> 		void mouseMoveOverTag (int tagId, int x, int y);
> 
> 	protected:
> 		void hoverEnterEvent (QGraphicsSceneHoverEvent * event);
> 		void hoverLeaveEvent (QGraphicsSceneHoverEvent * event);
> 		void hoverMoveEvent (QGraphicsSceneHoverEvent * event);
> 
> 	private:
> 		int tagId;
> };
> 
> class RDLDMap : public QGraphicsScene
> {
> 	Q_OBJECT
> ...
> 	public slots:
> 		void addTagItem (int tagId, QColor color);
> ...
> 	private:
> 		QMap <int, TagItem *>	tags;
> ...
> };
> 
> I tried it also without the isEmpty and only the contains, resulting
> into:
> if (tags.contains(tagId))
> 
> Program received signal SIGSEGV, Segmentation fault.
> 0x000000000041d606 in QMap<int, TagItem*>::findNode
> (this=0x706f746b73654483,
> akey=@0x7fffafdac2c4) at /usr/include/qt4/QtCore/qmap.h:424
> 424         QMapData::Node *cur = e;
> 
> and i tried the direct access:
> if (tags[tagId])
> 
> Program received signal SIGSEGV, Segmentation fault.
> 0x000000000041da58 in QMap<int, TagItem*>::detach
> (this=0x706f746b73654483)
> at /usr/include/qt4/QtCore/qmap.h:168
> 168         inline void detach() { if (d->ref != 1) detach_helper(); }
> 
> Thank you for your help in advance.
> 
> --
> To unsubscribe - send a mail to qt-interest-request@xxxxxxxxxxxxx with
> "unsubscribe" in the subject or the body.
> List archive and information: http://lists.trolltech.com/qt-interest/

N‹§²æìr¸›zÇu©šŠ[hªØµêÞÚÞçÚè–[^r(›­†éì®&Þ{azˍç-¢»ayºȬµªÜ+Þjwbú+™«b¢xm¶Ÿÿ–+-²Úè–[^r(›ú­Š{^­ë-
Message 3 in thread

Scott Aron Bloom wrote:
>Your RDLDMap is probably null or previously deleted.

I'd go further and say the object that contains RDLDMap has already been 
deleted. Simon did not show us the code that calls RDLDMap::addTagItem, 
which is where the problem is.

>> Program received signal SIGSEGV, Segmentation fault.
>> 0x000000000041da14 in QMap<int, TagItem*>::isEmpty
>> (this=0x706f746b73654483)

The "this" pointer above is obviously garbage. You don't have to 
understand the platform in question (looks like AMD64 Linux) to know it. 
All you have to do is see that it ends in "3", which is not a multiple of 
4.

With very, very few exceptions (and I can't think of any right now), Qt 
classes contain at least one pointer, which means they get the alignment 
requirement of pointers in that platform. Which is to say that the entire 
class must be aligned on 32- or 64-bit boundaries.

-- 
 [ signature omitted ]

Attachment: signature.asc
Description: This is a digitally signed message part.

Message 4 in thread

Thank you both
next time i go to bed first and ask the question later.

Thiago Macieira wrote:
> Scott Aron Bloom wrote:
> >Your RDLDMap is probably null or previously deleted.
>
> I'd go further and say the object that contains RDLDMap has already been
> deleted. Simon did not show us the code that calls RDLDMap::addTagItem,
> which is where the problem is.
>

I accidently created a new object of RDLDMap with the same name in the 
constructor instead of using the private of the class

> >> Program received signal SIGSEGV, Segmentation fault.
> >> 0x000000000041da14 in QMap<int, TagItem*>::isEmpty
> >> (this=0x706f746b73654483)
>
> The "this" pointer above is obviously garbage. You don't have to
> understand the platform in question (looks like AMD64 Linux) to know it.
> All you have to do is see that it ends in "3", which is not a multiple of
> 4.
I never heard of the ending in 3 instead of a multiple of 4, what is this for?

> With very, very few exceptions (and I can't think of any right now), Qt
> classes contain at least one pointer, which means they get the alignment
> requirement of pointers in that platform. Which is to say that the entire
> class must be aligned on 32- or 64-bit boundaries.

I do not understand what this has to do with my problem.


--
 [ signature omitted ]
Message 5 in thread

> Thank you both
> next time i go to bed first and ask the question later.
> 
> Thiago Macieira wrote:
> > Scott Aron Bloom wrote:
> > >Your RDLDMap is probably null or previously deleted.
> >
> > I'd go further and say the object that contains RDLDMap has already
> been
> > deleted. Simon did not show us the code that calls
> RDLDMap::addTagItem,
> > which is where the problem is.
> >
> 
> I accidently created a new object of RDLDMap with the same name in the
> constructor instead of using the private of the class
> 
> > >> Program received signal SIGSEGV, Segmentation fault.
> > >> 0x000000000041da14 in QMap<int, TagItem*>::isEmpty
> > >> (this=0x706f746b73654483)
> >
> > The "this" pointer above is obviously garbage. You don't have to
> > understand the platform in question (looks like AMD64 Linux) to know
> it.
> > All you have to do is see that it ends in "3", which is not a
> multiple of
> > 4.
> I never heard of the ending in 3 instead of a multiple of 4, what is
> this for?
> 
> > With very, very few exceptions (and I can't think of any right now),
> Qt
> > classes contain at least one pointer, which means they get the
> alignment
> > requirement of pointers in that platform. Which is to say that the
> entire
> > class must be aligned on 32- or 64-bit boundaries.
> 
> I do not understand what this has to do with my problem.

What he was saying, is if you analize the value of this,
(0x706.....4483) is that the memory address was clearly corrupt since
pointers must be aligned on a memory boundry... 

Allowing you to see that it was not a QT issue, but rather a memory
management issue on your part....

That said, Im glad you are on the right track to find your problem

Scott


--
 [ signature omitted ]
Message 6 in thread

Simon Schäfer wrote:

> I never heard of the ending in 3 instead of a multiple of 4, what is this for?
> 
>> With very, very few exceptions (and I can't think of any right now), Qt
>> classes contain at least one pointer, which means they get the alignment
>> requirement of pointers in that platform. Which is to say that the entire
>> class must be aligned on 32- or 64-bit boundaries.
> 
> I do not understand what this has to do with my problem.

Your pointer address was not aligned on a 4-byte boundary.
Most processors require that memory accesses be made to
addresses aligned to some "natural" multiple of a byte,
usually 4, else they execute an exception, which your
program eventually displays as a core dump.

So if you ever see a pointer ending in anything other
than a multiple of 4, there's a good chance that it
is invalid, and that something bad will happen if you
try to dereference it.

-- 
 [ signature omitted ]