Re: Is it time for an ABI/API breaking libxml++ 3.0 release?

[Murray Cumming <murrayc murrayc com>]

On Sat, 2015-09-05 at 12:21 +0200, Kjell Ahlstedt wrote:
> libxml++ uses the deprecated std::auto_ptr in some header files. It
> can't be replaced by std::unique_ptr without breaking ABI. See
> https://bugzilla.gnome.org/show_bug.cgi?id=753123.
>
> According to http://en.cppreference.com/w/cpp/memory, std::auto_ptr
> will be gone in C++17

I doubt that will actually happen, by the way.

> , meaning that we have to make an ABI breaking libxml++ release within
> a few years. IMO the sooner the better. If we make it pretty soon, the
> present libxml++-2.6.so can exist in parallel with a new libxml
> ++-3.0.so during several years while libxml++-2.6.so gradually becomes
> uninteresting. Application programmers will have enough time to adapt
> to the change.

Yes, an ABI-breaking parallel-installing version wouldn't bother anyone.
It's not such an important dependency as glibmm or libsigc++. Let's do
it. Feel free to start.
> Is 3.0 the right version number? Or shall it be 2.something as long as
> it's the C++ binding of libxml2?

3.0 seems fine.
> When we do make an ABI breaking release, we shall of course take the
> chance to make many other ABI and API breaking changes. There are TODO
> comments to consider, and other open bugs requesting ABI breaking
> fixes.
Great. It would be nice to take care of them.

Thanks.

-- 
Murray Cumming
murrayc murrayc com
www.murrayc.com