Shibboleth Developers

[Scott Cantor <>]

> This implies that all uses of dynamic_cast<> are subject to 
> this bug, eh?  That means I need to fix my code as well.

I think it's probably possible to use it in some contexts, mostly if
you're within the same body of code that built the object and would have
implemented the RTTI supporting code. It also appears to be safe in
executables that are totally linked in advance.

It seems to come up more in cases where you write one library that casts
an object implemented in a second library, so any casting of SAML
objects down would most likely break.

> Where is it safe and where is it not safe to use C-casts?

Any time you really know the pointer's real type, you're safe, barring
unforeseen errors. OTOH, if for example you have a collection of common
objects (say, oh, an assertion's statements) with no way to know what
type of statement each one is, there's no possible way to make it safe.
You simply act like you're in C, and you stick additional functions or
data in to tell you what the story is. That's all dynamic_cast and
typeinfo are anyway, but the compiler does it for you.

-- Scott

------------------------------------------------------mace-shib-design-+
For list utilities, archives, subscribe, unsubscribe, etc. please visit the
ListProc web interface at 

    http://archives.internet2.edu/

------------------------------------------------------mace-shib-design--