Martin Bochnig wrote: >Good and careful ABI-stability handling with SunCC: >http://developers.sun.com/sunstudio/articles/CC_abi/CC_abi_content.html >"Here an ABI, There an ABI ..." (article does not cover gcc/g++ at all, >only the Sun-C and C++ compilers cc/CC over time) > > >On the other hand g++: >ABI instability over the years (i.e. from 2.x to 3.x series), but >apparently a certain striving for future ABI stability > >http://www.techfak.uni-bielefeld.de/rechner/bulletin/volume29/1.html > > >Translation: > >Problems > >Just as with previous GCC-Releases, the C++-Compiler g++'s ABI has been >made incompatible again, so that all C++-programs and -libraries will >have to be recompiled with the g++3.1 once again. Though g++ does on all >platforms implement a new C++-ABI http://www.codesourcery.com/cxx-abi/ >now, that had initially been developed, to achieve interoperability >among different vendor's C++-Compilers on the Itanium architecture. >Hence it will at least be possible in the future, to stay >binary-compatible from g++ version to g++ version, if not even to >achieve interoperability with other vendor's C++-Compilers on >Non-Itanium-systems, if - and only if - those 3rd party compilers also >implement above new interface. However, this binary compatibility is not >yet provided until now, since the GNU C++-Library libstdc++'s interface >is still subject of potential change, from Major Release to Major >Release, and therefore can become incompatible again. But also on that >front is being worked on, so that the steady requirement of having to >recompile C++-code with g++ will soon be a thing of the past. > > >
And ..., believe it or not: I still like GCC3.x Due to the aforementioned reasons (two/three mails ago). And I'm not guessing here, but have delivered arguments and url references. >-Martin > _______________________________________________ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
