From: Ralf W. Grosse-Kunstleve (rwgk_at_[hidden])
Date: 2008-05-02 15:08:19
> In my opinion, this warning should be issued only if an attempt to
> call the non-virtual destructor of a type with virtual functions
> through a base class pointer is actually made, not simply because the
> possibility exists.
What's the disadvantage of adding the virtual destructors?
It certainly is a huge disadvantage if -Wall cannot be used anymore.
Until rev. 44954 everything we have compiled noise-free with -Wall
(heaps of sources, many different gcc versions).
I think the warning is useful. A library author cannot anticipate
all situations in which the library may be used. If someone wants
to derive from a class in a library and this triggers the smarter
warning you propose for the first time, it is probably too late.
If the library is part of the system headers it cannot be changed.
The user is stuck.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk