Re: dynamic_cast behavior differs from standard?

[David Mazieres <dm at reeducation-labor dot lcs dot mit dot edu>]

> From: Joe Buck <jbuck@Synopsys.COM>
> Date: Tue, 23 Feb 99 16:52:34 PST
> 
> The place to ask such question is on comp.std.c++, where you're likely
> to get answers from the people who actually wrote the standard.
> 
> The fact that you say that all the other compilers are doing exactly
> what egcs is doing suggest that it is your reading rather than the
> compiler that is wrong.  (Though of course it's possible that all the
> implementers are making the same mistake).

Consider my message a bug report.  I don't want to say "hey, go
implement this change" if there's any chance I could be wrong.  I am
pretty sure this is a bug in egcs, though--sure enough that I think
it's worth filing a bug report.  Since I'm not exactly 100% sure,
though, I think it's also fair for me to say "but I could be wrong.
Here's some language from the standard.  Don't consider this a bug
report if there is some other possible interpretation."  Maybe I
sounded too willing to accept the blame for misreading the standard
because I was trying to be polite.

At any rate, even if the standard can somehow be interpreted to say
that regular, compile-time, access control also applies to
dynamic_cast, I still think the example code I included indicates a
bug in egcs.  (I didn't get into this in the previous mail because I
don't know exactly how this interpretation of the standard would come
about).

I'm sending this to egcs because the egcs developers actually seem to
pay attention to bug reports.  If I thought that was true of other
compiler vendors, I'd send them the bug report too.  If I saw any
ambiguity in the standard, I would post to comp.std.c++, but I don't.
The standard seems pretty clear.  (The word "friend" does not even
appear anywhere in section 5).  Since all the non-egcs compilers I
have access to are probably pre-standard, I suspect this may have been
a late change to the standard.

David