Re: [osg-users] Strange problem with osgUtil::LineSegmentIntersector

Andreas Goebel Mon, 16 Apr 2007 14:37:55 -0700

Hi,

the problem was the std::vector copy-constructor. I am probably stillhaving problems with different c-runtimes somewhere, though I carefullytried to avoid this.

Using a reference instead of a copy of the intersection-vector made thecrash go away (hopefully for ever):


           if (LineIntersector->containsIntersections() ){

osgUtil::LineSegmentIntersector::Intersections& MyInts =LineIntersector->getIntersections();for(osgUtil::LineSegmentIntersector::Intersections::iterator it =MyInts.begin();

                   it != MyInts.end(); ++it)

MyIntersections.push_back(it->getWorldIntersectPoint() );

SetPosition(LineIntersector->getFirstIntersection().getWorldIntersectPoint());

Though the reference is better code at this place anyway I wonder how Ican get rid of those c-runtime problems.

For those who don´t know about this: If you are using differentc-runtimes (msvcrt.dll and libcmt.dll for instance) thenmemory-management doesn´t work between the dlls. In this case the memorywas allocated in osgUtil, and the std::vector copy-constructor tried toreuse it. Thus it had to be deallocated in my program (as I took acopy), which made it crash, probably because osgUtil used msvcrt.dll forthe allocation and I used libcmt.dll for the deallocation (or viceversa), I´ll dig into this.

Am I right in assuming that the whole osg is linking against libcmt(multi threaded c-runtime), or is this wrong?


Regards,

Andreas

Robert Osfield schrieb:

Hi Andreas,

Try adding adding LineSegmentIntersector::Intersection copy
constructor, assignment operator and destructor to see if this has any
effect, perhaps MSVC isn't creating these appropriately.

Robert.

On 4/16/07, Andreas Goebel <[EMAIL PROTECTED]> wrote:

Andreas Goebel schrieb:
>

>> MSVC allows you to generate debug info even in Release Mode; makesure

>> you set the Debug Information Format in the C++ project options, and
>> tell the Linker to generate Debugging output. That might produce more
>> useful information when you run in the debugger.
>>
>> HTH,
>>
>
> Thanks,
>
> that´s a good idea. Actually the program crashes here:
>
> MyIntersections.clear();
>        if (LineIntersector.containsIntersections() ){
>            osgUtil::LineSegmentIntersector::Intersections MyInts =
> LineIntersector.getIntersections();
>            for
> (osgUtil::LineSegmentIntersector::Intersections::iterator it =
> MyInts.begin();
>                it != MyInts.end(); ++it)
>
> MyIntersections.push_back(it->getWorldIntersectPoint() );
>

>SetPosition(LineIntersector.getFirstIntersection().getWorldIntersectPoint()

> );
>        } //Here is the crash
>        else
>            SetPosition(osg::Vec3(0, 0, 0) );
>
> in the destructor of osgUtil::LineSegmentIntersector::Intersection
> (the crash is where the Intersections - vector goes out of scope).
>
> I´ll have a look at that destructor, but maybe Robert has an idea
> about that, too.
Aha, there is no desctructor, it´s just a simple struct. HM.

Here is the call-sequenze of that crash:

     ntdll.dll!7c911230()
     [Unten angegebene Rahmen sind möglicherweise nicht korrekt und/oder
fehlen, keine Symbole geladen für ntdll.dll]
     ntdll.dll!7c97c943()
     ntdll.dll!7c97cd80()
     ntdll.dll!7c97df66()
     ntdll.dll!7c92056d()
     msvcr80.dll!78134b9f()
     msvcr80.dll!78134bbe()
     ntdll.dll!7c95a5d0()
     osgUtil.dll!009db7c7()
     ntdll.dll!7c9368ad()
     DreiDEdit.exe!memmove_s(void * dst=0x01574288, unsigned int
sizeInBytes=24, const void * src=0x018b6e40, unsigned int count=24)
Zeile 58 + 0xc Bytes    C
     DreiDEdit.exe!std::_Uninit_copy<double const *,double
*,std::allocator<double> >(const double * _First=0x018b6e40, const
double * _Last=0x018b6e58, double * _Dest=0x014f0000,
std::allocator<double> & __formal={...}, std::allocator<double> &
__formal={...}, std::allocator<double> & __formal={...})  Zeile 129 +
0xc Bytes    C++

DreiDEdit.exe!stdext::unchecked_uninitialized_copy<std::_Vector_const_iterator<double,std::allocator<double>

 >,double *,std::allocator<double>
 >(std::_Vector_const_iterator<double,std::allocator<double> >
_First=0.45205613970756531,
std::_Vector_const_iterator<double,std::allocator<double> >
_Last=6.9850276205963215e+107, double * _Dest=0x7c936ac8,
std::allocator<double> & _Al={...})  Zeile 674 + 0x1d Bytes    C++
     DreiDEdit.exe!std::vector<double,std::allocator<double>
 >::_Ucopy<std::_Vector_const_iterator<double,std::allocator<double> >
 >(std::_Vector_const_iterator<double,std::allocator<double> >
_First={...}, std::_Vector_const_iterator<double,std::allocator<double>
 > _Last=3.785200942584e-317#DEN, double * _Ptr=0x018b6c08)  Zeile 1105
+ 0x15 Bytes    C++
     ntdll.dll!7c9368ad()
     ntdll.dll!7c9206eb()
     msvcr80.dll!78134ce9()
     msvcr80.dll!7817ff7a()
     osgUtil.dll!009d8df9()
     osgUtil.dll!009d90d2()
     DreiDEdit.exe!free(void * pBlock=0x018b6c08)  Zeile 110    C
     DreiDEdit.exe!std::vector<double,std::allocator<double> >::_Tidy()
Zeile 1096 + 0x6 Bytes    C++

DreiDEdit.exe!osgUtil::LineSegmentIntersector::Intersection::~Intersection()

+ 0x20 Bytes    C++
 >    DreiDEdit.exe!GeoPoint::Realize()  Zeile 78    C++


Looks like a strange problem.

Regards,

Andreas
>
> Regards,
>
> Andreas
>
> _______________________________________________
> osg-users mailing list
> [email protected]
> http://openscenegraph.net/mailman/listinfo/osg-users
> http://www.openscenegraph.org/
>
>


_______________________________________________
osg-users mailing list
[email protected]
http://openscenegraph.net/mailman/listinfo/osg-users
http://www.openscenegraph.org/

_______________________________________________
osg-users mailing list
[email protected]
http://openscenegraph.net/mailman/listinfo/osg-users
http://www.openscenegraph.org/



_______________________________________________
osg-users mailing list
[email protected]
http://openscenegraph.net/mailman/listinfo/osg-users
http://www.openscenegraph.org/

Re: [osg-users] Strange problem with osgUtil::LineSegmentIntersector

Reply via email to