libstdc++ version bump.

Alexander S. Usov A.S.Usov at kvi.nl
Tue Jun 28 22:05:22 GMT 2005


Alexander S. Usov wrote:

> David O'Brien wrote:
> 
>> On Tue, Jun 28, 2005 at 06:44:02PM +0200, Michael Nottebrock wrote:
>>> On Tuesday, 28. June 2005 18:21, Scot Hetzel wrote:
>>> 
>>> > Are you sure that the problem is with libstdc++ and not the above
>>> > libraries?
>>> 
>>> Yes, it's a well known-fact. In fact there might have been a break of
>>> ABI backwards-compatibility between gcc 3.3 and 3.4, too, but I'm not
>>> sure.
>> 
>> There was.  3.3 could support either the C++ ABI of GCC 3.2, in which
>> there was an implementation bug; or the fixed ABI of 3.4.  I had hoped we
>> had gone with the fixed ABI rather than being G++ 3.2 compatible, but
>> that is what users wanted at the time.
> 
> Do I read it correctly that on 5.4 (with 3.4 in it) it is possible to
> convince lang/gcc33 to be binary compatible with system libraries?
> Do you have any link for it?

I think I have to clarify my question a bit:
I have a program which is a mix of c++ & fortran code. Fortran part is
compiled with lang/ifc, so to link it I use something like
ifort -o prog  for_lib.a cpp_lib.a -l.... -lstdc++

Now I have to use one more library, which is not (yet) compatible with 
gcc 3.4 and up. If it matters -- this library is luabind (luabind.sf.net)
and I don't really know of any alternatives to it (except tolua/tolua++
which are even more problematic). So to link my code now I have to rebuild
a bunch of other packages using gcc33 and link it like that:
ifort -o prog ...... /usr/local/lib/gcc-lib/......./libstdc++.a

Doing it this way is quite messy, provided that I have to build/use this
code time to time on linux machines where it should be linked in absolutely
different way.

So the question is if it possible to convince lang/gcc33 to produce code
binary compatible with system-wide gcc 3.4?

-- 
Best regards,
  Alexander.



More information about the freebsd-stable mailing list