This is the mail archive of the libstdc++@sourceware.cygnus.com mailing list for the libstdc++ project. See the libstdc++ home page for more information.
> It would help if I understood their meaning better. No problem with that. What's your question? (honor-std is documented, BTW). > In fact, I'd be glad to understand what part of the compiler/library > code is actually affected/concerned in supporting them. In general, the compiler is concerned with supporting the ABI (Application Binary Interface); the library is affected if this changes, as well as most programs are affected. > Also, I'd like to be able to understand the exact goal of the new-abi > project, in order either to contribute, or to make a proposal of mine. The goal (AFAICT) is to provide a means to distribute experimental features whose use would break binary compatibility. What these exact features are might change from snapshot to snapshot; I guess this is the reason why they are not documented. Currently, it is a combination of: - a new layout of base classes, saving space when you can, - vtable thunks, if they are not enabled by default, - the squangling algorithm, and - the activation of std:: as a true namespace. Soon, it might also cover a change in the vtable thunks which is not fully compatible with the current implementation. > /tmp/ccySiaaa.o(.text+0x1e): undefined reference to `ostream::operator<<(ostream &(*)(ostream &))' [...] > So, I still do something wrong... I'm somewhat lost here as to what exactly you are doing. I thought this is libstdc++ v3, where ostream should be a template, not a class. Also, the correct symbol should be std::ostream::operator<<(std::ostream &(*)(std::ostream &)) Please compile your programs with --save-temps, then analyse the object files using nm(1); do not use nm's demangling option. Then invoke nm on the library are linking with. This operator should start with __ls__. Without further information, I'd say you are getting the wrong header files. Please use --trace-includes or inspect the resulting .ii file to verify if that is the case. Hope this helps, Martin