[lug] ABI of RH's 2.96 g++

Tom Tromey tromey at redhat.com
Thu Jun 13 19:58:45 MDT 2002

>>>>> ">" == D Stimits <stimits at idcomm.com> writes:

>> I assume that the vtables and issues related to using C++ with dlopen
>> modules would be very close regardless of 2.96 or 3.x documents? It
>> sounds like the 3.x documentation is good for this purpose.

Yes.  The vtables and issues should be similar.  Be aware that there
is a lot more to the C++ ABI: exception handling, name mangling, and
RTTI at the very least.  The 3.x ABI is pretty well defined; there
should be a link from the gcc page to its specification.

[ dwarf-2 ]

>> This probably would not have any effect on the scheme for making
>> C++ work with dlopen modules if using 3.1 docs to figure out how
>> the RH 2.96 g++ and binutils work. Does that sound basically
>> correct?

Yes.  I just mentioned that since you said something about being
interested in debugging.  But for understanding what is going on,
dwarf-2 won't have any effect.

>> And do you by any chance happen to have a URL for something like a
>> tutorial on g++ internals interacting with dlopen style modules, or
>> general issues such as how vtables are created, and what would
>> break their use across modules?

Nope.  I have heard of problems affecting RTTI when you dlopen things
privately, but I haven't really tracked the discussion closely.  I'm
not really that heavy of a C++ user these days.


More information about the LUG mailing list