Re: Stabilising some tools only HVMOPs?

[Wei Liu <wei.liu2@citrix.com>]

On Thu, Feb 18, 2016 at 09:41:01AM -0700, Jan Beulich wrote:
> >>> On 18.02.16 at 17:28, <Ian.Jackson@eu.citrix.com> wrote:
> > Wei Liu writes ("Re: Stabilising some tools only HVMOPs?"):
> >> I think we come to the conclusion that these HVMOPs should be made
> >> stable. And to do so I'm going to introduce a __XEN_TOOLS_STABLE__ macro
> >> for them to distinguish from __XEN_TOOLS__.  And then libxendevicemodel
> >> will have -D__XEN_TOOLS_STABLE__  only.
> >> 
> >> Does this sound sufficient?
> > 
> > It would be better to rename -D__XEN_TOOLS__ too, to
> > -D__XEN_TOOLS_UNSTABLE.
> 
> Even if a minor one, this will create a compatibility problem for
> out of tree code including the headers: Their builds will all of
> the sudden break, until they figure they need to go and
> #define this new manifest symbol. Otoh maybe we would
> actually like to break their builds this way, to make them aware
> of the fact. In which case maybe __XEN_TOOLS__ should be
> retained for the stable portions?
> 

I think we should break their build but I also want to be a bit nicer.
So off the top of my head, we can have something like:

#if defined (__XEN_TOOLS__)
# error "NOTE: if you want to continue to build against unstable tools interface, use __XEN_TOOLS_UNSTABLE__ instead"
#endif

And place this in public headers that used to have __XEN_TOOLS__.

Wei.


> Jan
>