(not) breaking the ABI

Paolo Carlini pcarlini@suse.de
Wed May 25 13:46:00 GMT 2005


chris jefferson wrote:

> After a brief discussion on IRC, and recent discussion on the main gcc
> mailing list, I discovered it is perhaps harder to break the ABI that
> I thought it was :)

Indeed, from some of yours sentences I got the impression that you was
overly careful ;) Traditionally, this is also my own attitude, by the
way. But since breaking the ABi bothers a lot of people and we want to
keep on improving the library it's important to know exactly -the
maximum- we can do. Let's figure it out together in detail during the
next months.

> As far as I've been able to find out, things we go into so_6 cannot:
>
> a) turn a POD into a non-POD
> b) add or remove virtual member functions of classes/structs
> c) add or remove data members of classes
> d) Change a function so does something fundamentally different, as you
> don't know which the linker will choose to pull in (you can of course
> just give your new function a new signiture and delete the old one).
>
> Out of interest, is that it? I tried looking around for a specific
> guide to what breaks ABIs and while I could find some good guides for
> C, I couldn't find one for C++.

Besides

    http://gcc.gnu.org/onlinedocs/libstdc++/abi.html

you mean? Certainly the document can be improved but it's a good
starting point!

Paolo.



More information about the Libstdc++ mailing list