Fix two bugs in the serivce protocol specification

Fix two bugs in the serivce protocol specification

- Add a @VM definition.
- Correct the type of the elements field in an Instance.

R=turnidge@google.com

Committed: https://github.com/dart-lang/sdk/commit/ecb2c0b9f22c129f5688b35912988fe63849ce40

[Cutch, 2015-10-09 13:23:44]

johnmccutchan@google.com changed reviewers:
+ turnidge@google.com

[danrubel, 2015-10-09 14:38:46]

On 2015/10/09 13:23:51, Cutch wrote:

With the introduction of VMUpdate, does the definition of a "fixed id" need to
change?
Perhaps something like this? (changes denoted with *)

A fixed id will never expire *for the life of a VM object*, but the object it
refers to may be collected. The VM uses fixed ids for objects like scripts,
libraries, and classes. *When a VMUpdate event occurs, clients must discard any
cached fixed ids and call getVM to obtain a new VM object*

[Cutch, 2015-10-09 14:41:34]

On 2015/10/09 14:38:46, danrubel wrote:
> On 2015/10/09 13:23:51, Cutch wrote:
> 
> With the introduction of VMUpdate, does the definition of a "fixed id" need to
> change?
> Perhaps something like this? (changes denoted with *)
> 
> A fixed id will never expire *for the life of a VM object*, but the object it
> refers to may be collected. The VM uses fixed ids for objects like scripts,
> libraries, and classes. *When a VMUpdate event occurs, clients must discard
any
> cached fixed ids and call getVM to obtain a new VM object*

Right now a VMUpdate just means that the *name* of the VM changed.

[danrubel, 2015-10-09 18:34:14]

On 2015/10/09 14:41:34, Cutch wrote:
> On 2015/10/09 14:38:46, danrubel wrote:
> > On 2015/10/09 13:23:51, Cutch wrote:
> > 
> > With the introduction of VMUpdate, does the definition of a "fixed id" need
to
> > change?
> > Perhaps something like this? (changes denoted with *)
> > 
> > A fixed id will never expire *for the life of a VM object*, but the object
it
> > refers to may be collected. The VM uses fixed ids for objects like scripts,
> > libraries, and classes. *When a VMUpdate event occurs, clients must discard
> any
> > cached fixed ids and call getVM to obtain a new VM object*
> 
> Right now a VMUpdate just means that the *name* of the VM changed.

Great to know. Thanks!

[turnidge, 2015-10-12 16:51:19]

https://codereview.chromium.org/1396603004/diff/20001/runtime/vm/service/serv...
File runtime/vm/service/service.md (right):

https://codereview.chromium.org/1396603004/diff/20001/runtime/vm/service/serv...
runtime/vm/service/service.md:771: 
Document parens here.  Something like:

We also allow parentheses in type expressions.  This is useful when a property
is an _Array_ of multiple independent types:

    (PropertyType1|PropertyType2)[]

https://codereview.chromium.org/1396603004/diff/20001/runtime/vm/service/serv...
runtime/vm/service/service.md:1467: @Instance[]|Sentinel[] elements [optional];
This would mean that we either return a vector of Instances or a vector of
Sentinel.  Neither of those is quite right.

What we are really returning is:

  (@Instance|Sentinel)[]

[Cutch, 2015-10-12 17:40:47]

https://codereview.chromium.org/1396603004/diff/20001/runtime/vm/service/serv...
File runtime/vm/service/service.md (right):

https://codereview.chromium.org/1396603004/diff/20001/runtime/vm/service/serv...
runtime/vm/service/service.md:771: 
On 2015/10/12 16:51:19, turnidge wrote:
> Document parens here.  Something like:
> 
> We also allow parentheses in type expressions.  This is useful when a property
> is an _Array_ of multiple independent types:
> 
>     (PropertyType1|PropertyType2)[]

Done.

https://codereview.chromium.org/1396603004/diff/20001/runtime/vm/service/serv...
runtime/vm/service/service.md:1467: @Instance[]|Sentinel[] elements [optional];
On 2015/10/12 16:51:19, turnidge wrote:
> This would mean that we either return a vector of Instances or a vector of
> Sentinel.  Neither of those is quite right.
> 
> What we are really returning is:
> 
>   (@Instance|Sentinel)[]
> 

Done.

[turnidge, 2015-10-12 17:56:42]

Heads up Devon, this will break your parser.

[turnidge, 2015-10-12 17:56:47]

lgtm

[Cutch, 2015-10-12 18:03:53]

Committed patchset #3 (id:40001) manually as
ecb2c0b9f22c129f5688b35912988fe63849ce40 (presubmit successful).

[devoncarew, 2015-10-12 18:08:09]

On 2015/10/12 17:56:42, turnidge wrote:
> Heads up Devon, this will break your parser.

Thanks! Will update. The change to parans makes sense -