[Insight-developers] Question about delete on VC++
Miller, James V (Research)
millerjv@crd.ge.com
Mon, 30 Sep 2002 16:16:35 -0400
The problem in this case is that foo
1. Had not functions or ivars, just typedefs
2. The compiler generated a default constructor/destructor for foo.
3. Subclasses of foo defined the destructor to be virtual
i.e.
-----foo.cxx----
struct foo
{
};
struct bar : public foo
{
virtual ~bar();
};
> -----Original Message-----
> From: Bill Hoffman [mailto:bill.hoffman@kitware.com]
> Sent: Monday, September 30, 2002 4:11 PM
> To: Miller, James V (Research); 'Joshua Cates'
> Cc: 'Jisung Kim'; insight-dev-list
> Subject: RE: [Insight-developers] Question about delete on VC++
>
>
> I think something else is going on here...
>
> -----foo.cxx----
>
> struct foo
> {
> virtual int b();
> ~foo();
> };
>
>
> struct bar : public foo
> {
> ~bar();
> };
>
> Linux:
> [hoffman@ringworld hoffman]$ c++ -Wall -c foo.cxx
> foo.cxx:5: warning: `struct foo' has virtual functions but
> non-virtual destructor
> foo.cxx:11: warning: `struct bar' has virtual functions but
> non-virtual destructor
>
> SGI:
> rolle:/rolle/insight % CC -fullwarn -LANG:std -c foo.cxx
> cc-1375 CC: REMARK File = foo.cxx, Line = 8
> The destructor for base class "foo" is not virtual.
>
> struct bar : public foo
> ^
>
> cygwin:
> bash-2.05$ c++ -Wall foobar.cxx
> foobar.cxx:4: destructor `f' must match class name `foo'
> foobar.cxx:5: warning: `struct foo' has virtual functions but
> non-virtual destructor
> foobar.cxx:11: warning: `struct bar' has virtual functions
> but non-virtual destructor
> bash-2.05$ c++ -v
> Reading specs from /usr/lib/gcc-lib/i686-pc-cygwin/2.95.3-5/specs
> gcc version 2.95.3-5 (cygwin special)
>
>
>
>
>
> At 03:53 PM 9/30/2002 -0400, Miller, James V (Research) wrote:
> >I think the warning that the compiler reports is for a
> >class with a non-virtual destructor and and a subclass
> >with a virtual destructor.
> >
> >The hierarchy here used STRUCTS as opposed to classes.
> >So I guessing the compiler didn't warn about structs.
> >
> >
> >
> >> -----Original Message-----
> >> From: Joshua Cates [mailto:cates@sci.utah.edu]
> >> Sent: Monday, September 30, 2002 3:48 PM
> >> To: Bill Hoffman
> >> Cc: Miller, James V (Research); 'Jisung Kim'; insight-dev-list
> >> Subject: RE: [Insight-developers] Question about delete on VC++
> >>
> >>
> >> One of the builds warns about this. I think IRIX.
> >>
> >> Josh.
> >>
> >> ______________________________
> >> Josh Cates
> >> School of Computer Science
> >> University of Utah
> >> Email: cates@sci.utah.edu
> >> Phone: (801) 587-7697
> >> URL: www.cs.utk.edu/~cates
> >>
> >>
> >> On Mon, 30 Sep 2002, Bill Hoffman wrote:
> >>
> >> > Isn't this a warning on some compilers?
> >> > (Class with virutal table has non-virutal destructors?)
> >> >
> >> >
> >> > -Bill
> >> >
> >> >
> >> > At 03:41 PM 9/30/2002 -0400, Miller, James V (Research) wrote:
> >> > >I ran the test through Purify. This identified that a leak
> >> > >occurred. The I spent several hours trying to find why your
> >> > >memory was not being deallocated. Eventually, I put print
> >> > >statements in each node type's destructor to make sure they
> >> > >were being called. As it turned out, the destructors were
> >> > >never being called because the nodes in the kdtree
> store pointers
> >> > >to a kdnode and kdnode did not have a destructor defined. This
> >> > >meant that the compiler provided its own destructor which by
> >> > >default is non-virtual.
> >> > >
> >> > >Jim
> >> > >
> >> > >
> >> > >
> >> > >> -----Original Message-----
> >> > >> From: Jisung Kim [mailto:bahrahm@yahoo.com]
> >> > >> Sent: Monday, September 30, 2002 2:16 PM
> >> > >> To: Miller, James V (Research); insight-dev-list
> >> > >> Subject: RE: [Insight-developers] Question about
> delete on VC++
> >> > >>
> >> > >>
> >> > >> Thanks, Jim.
> >> > >>
> >> > >> Could you tell me how did you found the memory leaks?
> >> > >> I was experimenting with _CrtDumpMemoryLeaks()
> >> > >> function. It would be helpful in future.
> >> > >>
> >> > >> Thank you very much,
> >> > >>
> >> > >> Jisung.
> >> > >>
> >> > >> --- "Miller, James V (Research)" <millerjv@crd.ge.com>
> >> > >> wrote:
> >> > >> > Jisung,
> >> > >> >
> >> > >> > I just fixed the memory leaks in your KdTree.
> >> > >> >
> >> > >> > The base kd-node class did not have a "virtual"
> >> > >> > destructor. When the
> >> > >> > tree was being deleted, the destructors for "real"
> >> > >> > kd-nodes was not
> >> > >> > being called.
> >> > >> >
> >> > >> > Jim
> >> > >> >
> >> > >> >
> >> > >> >
> >> > >> > > -----Original Message-----
> >> > >> > > From: Jisung Kim [mailto:bahrahm@yahoo.com]
> >> > >> > > Sent: Friday, September 27, 2002 1:53 PM
> >> > >> > > To: Jisung Kim; Miller, James V (Research);
> >> > >> > insight-dev-list
> >> > >> > > Subject: RE: [Insight-developers] Question about
> >> > >> > delete on VC++
> >> > >> > >
> >> > >> > >
> >> > >> > > The print out result shows exact match between
> >> > >> > > addresses of pointer (to nodes) and the order.
> >> > >> > >
> >> > >> > >
> >> > >> > > --- Jisung Kim <bahrahm@yahoo.com> wrote:
> >> > >> > > > In a same thread.
> >> > >> > > >
> >> > >> > > > I will do printing out address of nodes in
> >> > >> > creation
> >> > >> > > > and deletion in VC++ too. And let you know the
> >> > >> > > > result.
> >> > >> > > >
> >> > >> > > > --- "Miller, James V (Research)"
> >> > >> > > > <millerjv@crd.ge.com>
> >> > >> > > > wrote:
> >> > >> > > > > Are those in different threads?
> >> > >> > > > >
> >> > >> > > > >
> >> > >> > > > >
> >> > >> > > > > > -----Original Message-----
> >> > >> > > > > > From: Jisung Kim [mailto:bahrahm@yahoo.com]
> >> > >> > > > > > Sent: Friday, September 27, 2002 12:48 PM
> >> > >> > > > > > To: Miller, James V (Research);
> >> > >> > insight-dev-list
> >> > >> > > > > > Subject: RE: [Insight-developers] Question
> >> > >> > about
> >> > >> > > > > delete on VC++
> >> > >> > > > > >
> >> > >> > > > > >
> >> > >> > > > > > Yes, the tree was created by the
> >> > >> > > > > > itkWeightedCenteroidKdTreeGenerator and the
> >> > >> > > > > deletion
> >> > >> > > > > > logic is inside of the itkKdTree.
> >> > >> > > > > >
> >> > >> > > > > > Is it a problem?
> >> > >> > > > > >
> >> > >> > > > > > Thanks,
> >> > >> > > > > >
> >> > >> > > > > >
> >> > >> > > > > > --- "Miller, James V (Research)"
> >> > >> > > > > <millerjv@crd.ge.com>
> >> > >> > > > > > wrote:
> >> > >> > > > > > > Another thought. Is the data allocated in
> >> > >> > one
> >> > >> > > > > > > thread and destroyed
> >> > >> > > > > > > in another thread?
> >> > >> > > > > > >
> >> > >> > > > > > >
> >> > >> > > > > > >
> >> > >> > > > > > > > -----Original Message-----
> >> > >> > > > > > > > From: Miller, James V (Research)
> >> > >> > > > > > > > Sent: Friday, September 27, 2002 12:07
> >> > >> > PM
> >> > >> > > > > > > > To: 'Jisung Kim'; insight-dev-list
> >> > >> > > > > > > > Subject: RE: [Insight-developers]
> >> > >> > Question
> >> > >> > > > > about
> >> > >> > > > > > > delete on VC++
> >> > >> > > > > > > >
> >> > >> > > > > > > >
> >> > >> > > > > > > > Strange. If we were building Dlls, I
> >> > >> > would
> >> > >> > > > > say
> >> > >> > > > > > > that maybe we were
> >> > >> > > > > > > > allocating nodes in one DLL and
> >> > >> > deallocating
> >> > >> > > > > them
> >> > >> > > > > > > in another DLL.
> >> > >> > > > > > > > (I think each DLL has its own heap
> >> > >> > > > > management).
> >> > >> > > > > > > But since we are
> >> > >> > > > > > > > not building DLLs, then that is not the
> >> > >> > > > > problem.
> >> > >> > > > > > > >
> >> > >> > > > > > > > Have you run with VC the version that
> >> > >> > prints
> >> > >> > > > > out
> >> > >> > > > > > > all the node
> >> > >> > > > > > > > allocation and deletions?
> >> > >> > > > > > > >
> >> > >> > > > > > > > Is there anything "inside" a node that
> >> > >> > is
> >> > >> > > > not
> >> > >> > > > > > > being deleted?
> >> > >> > > > > > > > Perhaps a
> >> > >> > > > > > > > subitem is missing a destructor that
> >> > >> > needs
> >> > >> > > > to
> >> > >> > > > > do
> >> > >> > > > > > > something
> >> > >> > > > > > > > that another
> >> > >> > > > > > > > compiler is providing by default.
> >> > >> > > > > > > >
> >> > >> > > > > > > >
> >> > >> > > > > > > >
> >> > >> > > > > > > >
> >> > >> > > > > > > > Friday, September 27, 2002 10:16 AM
> >> > >> > > > > > > > > To: insight-dev-list
> >> > >> > > > > > > > > Subject: [Insight-developers] Question
> >> > >> > > > about
> >> > >> > > > > > > delete on VC++
> >> > >> > > > > > > > >
> >> > >> > > > > > > > >
> >> > >> > > > > > > > > Hi.
> >> > >> > > > > > > > >
> >> > >> > > > > > > > > I have a question for memory
> >> > >> > deallocation
> >> > >> > > > in
> >> > >> > > > > > > VC++.
> >> > >> > > > > > > > >
> >> > >> > > > > > > > > Currently, my
> >> > >> > > > > itkKdTreeBasedKmeansEstimatorTest
> >> > >> > > > > > > is
> >> > >> > > > > > > > > leaking memory on VC++ machine. It
> >> > >> > seems
> >> > >> > > > > like
> >> > >> > > > > > > that the
> >> > >> > > > > > > > > some of the k-d tree nodes has not
> >> > >> > been
> >> > >> > > > > deleted.
> >> > >> > > > > > > > > However, On linux, it didn't have any
> >> > >> > > > leak.
> >> > >> > > > > I
> >> > >> > > > > > > traced
> >> > >> > > > > > > > > the creation and deletion of nodes by
> >> > >> > > > print
> >> > >> > > > > out
> >> > >> > > > > > > the
> >> > >> > > > > > > > > address of each nodes. And the result
> >> > >> > > > shows
> >> > >> > > > > > > perfect
> >> > >> > > > > > > > > match between creation and deletion.
> >> > >> > > > > > > > >
> >> > >> > > > > > > > > Is there anything more than calling
> >> > >> > > > "delete"
> >> > >> > > > > to
> >> > >> > > > > > > delete
> >> > >> > > > > > > > > an object. I think I read an article
> >> > >> > from
> >> > >> > > > MS
> >> > >> > > > > > > about
> >> > >> > > > > > > > > using delete and also assign zero to
> >> > >> > the
> >> > >> > > > > pointer
> >> > >> > > > > > > to
> >> > >> > > > > > > > > properly delete object in a long time
> >> > >> > ago.
> >> > >> > > > > Is it
> >> > >> > > > > > > the
> >> > >> > > > > > > > > case? Should I assign zero to the
> >> > >> > pointer
> >> > >> > > > > after
> >> > >> > > > > > > > > calling delete?
> >> > >> > > > > > > > >
> >> > >> > > > > > > > > Thank you,
> >> > >> > > > > > > > >
> >> > >> > > > > > > > >
> >> > >> > > > > > > > >
> >> > >> > > > > > > > > =====
> >> > >> > > > > > > > > Jisung Kim
> >> > >> > > > > > > > > bahrahm@yahoo.com
> >> > >> > > > > > > > > 106 Mason Farm Rd.
> >> > >> > > > > > > > > 129 Radiology Research Lab., CB# 7515
> >> > >> > > > > > > > > Univ. of North Carolina at Chapel Hill
> >> > >> > > > > > > > > Chapel Hill, NC 27599-7515
> >> > >> > > > > > > > >
> >> > >> > > > > > > > >
> >> > >> > > > > > >
> >> > >> > > > >
> >> > >> > __________________________________________________
> >> > >> > > > > > > > > Do you Yahoo!?
> >> > >> > > > > > > > > New DSL Internet Access from SBC &
> >> > >> > Yahoo!
> >> > >> > > > > > > > > http://sbc.yahoo.com
> >> > >> > > > > > > > >
> >> > >> > > > >
> >> > >> > _______________________________________________
> >> > >> > > > > > > > > Insight-developers mailing list
> >> > >> >
> >> > >> === message truncated ===
> >> > >>
> >> > >>
> >> > >> =====
> >> > >> Jisung Kim
> >> > >> bahrahm@yahoo.com
> >> > >> 106 Mason Farm Rd.
> >> > >> 129 Radiology Research Lab., CB# 7515
> >> > >> Univ. of North Carolina at Chapel Hill
> >> > >> Chapel Hill, NC 27599-7515
> >> > >>
> >> > >> __________________________________________________
> >> > >> Do You Yahoo!?
> >> > >> Yahoo! Health - Feel better, live better
> >> > >> http://health.yahoo.com
> >> > >>
> >> > >_______________________________________________
> >> > >Insight-developers mailing list
> >> > >Insight-developers@public.kitware.com
> >> > >http://public.kitware.com/mailman/listinfo/insight-developers
> >> >
> >> > _______________________________________________
> >> > Insight-developers mailing list
> >> > Insight-developers@public.kitware.com
> >> > http://public.kitware.com/mailman/listinfo/insight-developers
> >> >
> >>
> >_______________________________________________
> >Insight-developers mailing list
> >Insight-developers@public.kitware.com
> >http://public.kitware.com/mailman/listinfo/insight-developers
>