Soliciting help with libstdc++ and exceptions

[Anthony Green]

Martin wrote:
> First, why could it ever happen that one copy of libgcc.a was compiled
> with --enable-threads, and the other was not?

I don't believe in general that Linux distributors configure with
--enable-threads.  This means that most of the system's shared
libraries will contains bits of libgcc.a that are not compatible with
one configured with --enable-threads (assuming you build your own).

> More interesting, what is the specific problem that is causing the
> bug? AFAICT, none of the signatures involved depend on
> --enable-threads (including the layout of objects passed as
> parameters). All stuff that depends on --enable-threads is static in
> frame.c.

Yes, that should all stay the same.  I actually don't know the answer.
To really find out, I think what you should do is look at libgcc
today, and look at the libgcc that was used to build the conflicting
shared library (on my system, that would probably be 2.7.*).

AG

-- 
Anthony Green                                               Cygnus Solutions
                                                       Sunnyvale, California