Avoid crash

[P. Christeas]

Dominik Strasser wrote:
> I had a crash where This->pTypeInfo was 0.
> I didn't investigate the real culprit.
>
> Changelog: avoid crash on NULL pointer

I read your patch; I didn't investigate it either.
I think you should put the test inside the ITypeInfo_Release() fn. Your patch 
may solve this case, but what happens when that fn is mis-called from 
somewhere else? This is a general matter of programming style. Moreover, you 
could add some TRACE or warning point.

There was a conversation on what to do in such cases (where we could handle a 
NULL, but M$ may not). I suggest: warn and try to go on..