[Ns-developers] How to Deal With Interface Revisions (was RE:Learning Bridge has been merged)

[craigdo@ee.washington.edu]

> I am OK with your proposal, assuming deprecated APIs are not 
> removed for
> about one year.  I think achieving one year of relative 
> stability (where it
> doesn't compromise API cleanliness) is a good compromise between API
> stability and maintainability.  Maintainers, do not despair, 
> one year passes
> very quickly...
> 
> Now I am going to preempt Mathieu's likely reply and say that the big
> problem, in this case, is that I do not know how to make GCC emit a
> deprecation warning for a subclass of NetDevice that doesn't 
> override the
> virtual methods that will become pure in the future.  So this 
> is the only
> problem I can see to that approach.  Unfortunate problem, I 
> must add... :-(

Yup.  I thought that a deprecation warning could be emitted when the code
implementing the deprecated virtual method providing compatibility is
compiled.  This warning would hopefully trigger a user to go and read our
API readme to find what this means -- what is going to happen, and when.  If
this user maintains a net device she could go in at her leisure and
implement the required method and forget about the warning.

It would be nice to be able to emit user-defined warning text ...

-- Craig