Re: 4.22.0 with MSN still segfaulting on OpenBSD

With carefully-arranged electrons, Aaron Spanik wrote:
... snip ... 

... lots of template gunk ...

> Program received signal SIGSEGV, Segmentation fault.
> 648           _M_root() const { return (_Link_type&)
> _M_header->_M_parent; } 
> 
> (gdb) bt 
> 
> #0  std::_Rb_tree<std::string, std::pair<std::string const, void
> (MSN::Connection::*)(std::vector<std::string,
> std::allocator<std::string> >&, std::string, std::string)>,
> std::_Select1st<std::pair<std::string const, void
> (MSN::Connection::*)(std::vector<std::string,
> std::allocator<std::string> >&, std::string, std::string)> >,
> std::less<std::string>, std::allocator<std::pair<std::string const,
> void (MSN::Connection::*)(std::vector<std::string,
> std::allocator<std::string> >&, std::string, std::string)> >
> >::lower_bound(std::string const&) (this=0x0, __k=@0xfffffffffffe7490) at stl_tree.h:648 

This is a problem with static initialization -- not an STL problem, per
se, except for the fact that the STL expects the referenced static member of
MSN::Connection to be initialized before this algorithm is executed.

The same problem exists with Mac OS X on powerpc. I'm guessing that deep
in the bowels of the compiler, the different architectures imply
different static initialization ordering, and the version in centericq
is unstable on non-x86.

The design of the lib changes significantly between the versions; this
problem will probably go away with the new libmsn. Integrating it,
however, may prove to be a little challenging.

d

-- 
David W. Harks <dave-Ax18E/[email protected]>

_______________________________________________
Cicq mailing list
Cicq-xGejAJT2w6wWP6gT/[email protected]
http://mailman.linuxpl.org/mailman/listinfo/cicq
Questions? Check the FAQ first: http://centericq.de/faq/