config listener API ...

[Michael Stahl]

On 11.10.2015 10:12, Noel Grandin wrote:
> On 10 October 2015 at 15:41, Bjoern Michaelsen
> <bjoern.michaelsen at canonical.com> wrote:
>> And if something like boost::signal2 is awesome beyond believe (which is --
>> given Michaels hint about LOCs added at least open to debate), we should first
>> make our least broken implementation wrap or subclass that as a migration path.
> 
> Given that we have lambda's and std::function available now, it should
> be straightforward to convert our existing stuff (like tools/link.hxx)
> to these without the include overhead of boost::signal

boost::signals actually has a bunch of extra features that all of the
existing observers lack, such as groups and priorities, and the question
is whether the benefit of these extra features outweighs the obvious
compile time cost of boost::signals.

in the one case where we actually (AFAICR) use the very advanced feature
of multiple observers on one signal, it was possible to limit the header
interface to just std::function (see commit
20bd0a2ee9ed899ea542c2de08efda243dbef446).