RE: [Vxl-maintainers] Memory leaks in vipl

[Ian S. <ian...@st...>]

> From: Andrew Fitzgibbon [mailto:aw...@ro...]
> What is the status of Manchester image handling?

About the only advantage mil/algo has over vipl is that it doesn't leak
memory!

Everything of ours is specialised for fully in-memory images and uses the
step-multiplication style of indexing that mil provides. This means that it
doesn't work with vil.
We also don't have much in the way of algorithms - I'm just writing/testing
a 5-tap Gaussian smoother now.

> [the memory leaks] Looks like a fairly fundamental stumbling block to
me....

I wondered if many of the uses of smart pointers could be replaced with
ordinary pointers, and explicit deletes where necessary?

Ian.

> -----Original Message-----
> Sent: Wednesday, June 05, 2002 4:00 PM
> To: Ian Scott
> Subject: RE: [Vxl-maintainers] Memory leaks in vipl
>
>
>
>
>
>
> > -----Original Message-----
> > From: vxl...@li...
> > [mailto:vxl...@li...] On
> > Behalf Of Ian Scott
> > Sent: 05 June 2002 14:58
> > To: Vxl-maintainers (E-mail)
> > Subject: [Vxl-maintainers] Memory leaks in vipl
> >
> >
> > I would like to withdraw my previous suggestion of promoting
> > vipl and vepl
> > to full VXL status.
> >
> > vipl is full of memory leaks. These seem largely to be
> > connected with the
> > use of vbl_smart_ptr of objects that do not reference count.
> > The explicit
> > versions of vbl_smart_ptr<vipl_section_contatainer<T>
> > >::ref(v) etc merely
> > set v to 0, but don't delete the section. I can't see any
> > easy way of fixing
> > these problems.
> > Anyone else know how?
> >
> > Ian.
> >
> > I include the purify output of
> > 'vipl_test_driver vipl_test_gaussian_convolution'
> >
> > > -----Original Message-----
> > > From: Ian Scott
> > > Sent: Monday, May 27, 2002 11:59 AM
> > > To: Vxl-maintainers (E-mail)
> > > Subject: [Vxl-maintainers] Promote vipl and vepl to full VXL?
> > >
> > >
> > > Having played around with tbl recently, I was pleasantly
> > > surprised to find
> > > out how easy it was to use, despite its intimidating
> complexity. In
> > > particular, I found it really easy to extend. API's worked as
> > > their names
> > > suggested, and the extension route was straightforward.
> > >
> > > The only thing missing is a good overview for the VXL book.
> > > Is anyone aware
> > > of any overview documentation for vipl or vepl? It need
> not be that
> > > detailed - I was able to understand most of the
> > > implementation details from
> > > the doxygen comments in a few hours.
> > >
> > > Subject to some documentation appearing, I would nominate
> > > vipl and vepl for
> > > promotion to full VXL libraries.
> > >
> > > Ian.
> > >
> > >
> > > _______________________________________________________________
> > >
> > > Don't miss the 2002 Sprint PCS Application Developer's Conference
> > > August 25-28 in Las Vegas --
> > http://devcon.sprintpcs.com/adp/index.cfm
> > >
> > >
> > _______________________________________________
> > > Vxl-maintainers mailing list
> > > Vxl...@li...
> > > https://lists.sourceforge.net/lists/listinfo/vxl-maintainers
> > >
> >
>