[Insight-developers] Weird AutoPointer bug

[kent williams]

I was investigating this bug: http://public.kitware.com/Bug/view.php?id=1033
about memory leaks in the SimplexMesh class, and determined that the
particular bug described is no longer reproducable.  So I'm marking that bug
fixed.

There is a different bug in one of the SimplexMesh tests, that is really
curious. I logged a new bug about it:

http://public.kitware.com/Bug/view.php?id=11360

Valgrind reports leaks, but if I explicitly assign 0 to the SmartPointers to
the objects valgrind thinks are leaking (which forces the deletion of the
objects pointed to), the valgrind leaks disappear.

This makes absolutely no sense, since the SmartPointers will go out of scope
at the end of the test function, so there should be no need to set them to 0
before they go out of scope.

So I'm baffled.