Comment 17 for bug 1286611

I would like to add another view on such implementation details.

My small source code examples show only that specific destructors were not called after a exception was thrown. The tested behaviour fits to rules of the C++ programming language.
http://www.parashift.com/c++-faq/overview-dtors.html

I dare to point out this: How important is it for you to give each object the chance to perform its last task before deletion?
http://www.cprogramming.com/tutorial/constructor_destructor_ordering.html

Can it be that the value of a corresponding resource release operation is occasionally underestimated?
(Pointers will not be reset to the safe value "nullptr" for example which might result in unwanted effects if affected pointers were passed around in the software for other uses.)