Virtual Base Bug

[H.J. Lu]

> 
> > Anyway, an ABI change is planned for GCC in the not-so-distant future,
> > and, if nobody finishes Martin's fix for the vtable problem by then,
> > that would be a good moment to switch back to the slower yet fully
> > correct implementation.  I.e., since we're going to have to face a
> > large binary incompatibility problem, let's face it just once.
> 
> Such an ABI change would affect glibc as well, because of the shared
> libio code. Breaking the ABI for C++ is something that C++ users are
> used to. Breaking the C ABI is something that results in public
> uprise. As for the fix: I definitely want to complete it, it just gets
> delayed again and again.
> 

It seems that fixing this Linux specific bug is not the priority for
the most of gcc developers. I am sad to see Martin's patch never go
anywhere. If it is what has to take, I am willing to find sources to
make a Linux version of gcc which includes Martin's patch for this
"serious" bug. I will find a CVS server and a web server for the
project. Martin, you will have full access to the server. I will try
to keep it in sync with the official gcc. It is not the first time
for Linux. I hope it will be the last. It will be better if we don't
have to do it.

Thanks.

-- 
H.J. Lu (hjl@gnu.org)