Re: C++ PATCH: fix PR 10891

[Gabriel Dos Reis <gdr at cs dot tamu dot edu>]

Nathan Sidwell <nathan@codesourcery.com> writes:

| Gabriel Dos Reis wrote:
| > The issue with this PR is that we were accepting codes using
| > dynamic_cast while they specific -fno-rtti at the same time.  We then
| > produce object codes that crash at runtime.  If -fno-rtti, one can't
| > use dynamic_cast.  This patch makes us report the diagnostic at
| > compile-time.
| 
| > +   /* Use of dynamic_cast when -fno-rtti is a disaster waiting to happen.
| > +      See PR C++/10891.  Reject.  */
| 
| We don't usually refer to PR's in the compiler itself.

I believe you.  I saw references to PR in other parts of the compiler,
as witnesses of bugs we used to have, not as a susbitute for
documentation.  That was the intent here.

| the comment should be self contained possibly with reference to the relevent
| section of the appropriate standard. 

I'm all for self-contained comments and citation to appropriate part
of the standard.  Removing the reference to the PR does not make the
comment less self-contained.  It serves here as a witness of our past
failure. 

| In this case it's pretty obvious  that dynamic_cast requires rtti :)

Fully agreed. :-)

-- Gaby