treat void as a top type, analyzer fix for #28763

treat void as a top type, analyzer fix for #28763

void is treated as a top type in the type system.
(along with Object, dynamic, and FutureOr< any top type >)

[Jennifer Messerly, 2017-02-17 03:38:03]

Description was changed from

==========
fix analyzer side of #28763
void is treated as top in the type system
==========

to

==========
treat void as a top type, analyzer fix for #28763

void is treated as a top type in the type system.
(along with Object, dynamic, and FutureOr< any top type >)
==========

[Jennifer Messerly, 2017-02-17 03:38:03]

jmesserly@google.com changed reviewers:
+ leafp@google.com

[Leaf, 2017-02-17 18:12:58]

On 2017/02/17 03:38:53, Jennifer Messerly wrote:

I think we should hold off on this until we have a draft spec from Lasse (which
I would expect will be shared with the team shortly).  There's a couple of
things in here that I'm not sure match up exactly with the current proposal, and
I'd rather avoid hysteresis unless there's a good reason to do this now.

[Jennifer Messerly, 2017-02-17 18:20:51]

On 2017/02/17 18:12:58, Leaf wrote:
> On 2017/02/17 03:38:53, Jennifer Messerly wrote:
> 
> I think we should hold off on this until we have a draft spec from Lasse
(which
> I would expect will be shared with the team shortly).  There's a couple of
> things in here that I'm not sure match up exactly with the current proposal,
and
> I'd rather avoid hysteresis unless there's a good reason to do this now.

sounds good. Did we jump the gun implementing this in DDC?

[Leaf, 2017-02-17 20:30:30]

On 2017/02/17 18:20:51, Jennifer Messerly wrote:
> On 2017/02/17 18:12:58, Leaf wrote:
> > On 2017/02/17 03:38:53, Jennifer Messerly wrote:
> > 
> > I think we should hold off on this until we have a draft spec from Lasse
> (which
> > I would expect will be shared with the team shortly).  There's a couple of
> > things in here that I'm not sure match up exactly with the current proposal,
> and
> > I'd rather avoid hysteresis unless there's a good reason to do this now.
> 
> sounds good. Did we jump the gun implementing this in DDC?

No, I'm fine with the DDC runtime being ahead of the analyzer implementation.  I
think the current DDC semantics are compatible with Lasse's proposal.  Basically
at runtime void is Object (modulo pretty-printing), statically there are some
additional restrictions.  Similar treatment to dynamic in that way: Object +-
some static warnings.