stack-allocated Window subclasses

[Michael Meeks]

Hi Noel,

On Tue, 2014-09-23 at 14:52 +0200, Noel Grandin wrote:
> Otherwise we should just standardise on boost::intrusive_ptr and remove SvRef.

	Yep - my concern is only that boost::intrusive_ptr< Window > - is some
horribly long horror =) and that having some semi-standard means of
having widget refs that are not so grim might be nicer:

	WindowRef xRef;
vs.	boost::intrusive_ptr< Window > xRef;
vs.	vcl::ref< Window > xRef;

	etc. but would love to hear other people's views; in general, I think
fewer tokens for the mind to swallow when reading the bulk of the code
is a good thing.

> This should be fixable if it doesn't already work - myself and sberg
> already did this for rtl::Reference.

	Nice =)

> Mostly I guess we need to figure out where and when to destroy them -
> is it in the destructor, or in some other method specific to this task?

	Of course, the ideal is to migrate these stuff incrementally. To do
that, I'd suggest we use the new intrusive_ptr and destroy the Window in
the normal way at the end; but that we slowly migrate all explicit
'delete' calls to be virtual dispose calls, and split out reference
releasing / cyclic references later.

	My hope is that all these non-heap Window's are leafs in the hierarchy
anyhow and/or that we can clean them up without worrying about a proper
'dispose' mechanism just now.

	Does that make sense ?

	ATB,

		Michael.

-- 
 michael.meeks at collabora.com  <><, Pseudo Engineer, itinerant idiot