Ensure PrintToJSONStream can be called on all heap objects without crashing.

Ensure PrintToJSONStream can be called on all heap objects without crashing.

R=johnmccutchan@google.com

Committed: https://code.google.com/p/dart/source/detail?r=34271

[koda, 2014-03-21 18:07:27]

FYI, here are the two other issues that were uncovered by this unit test:
https://codereview.chromium.org/207303002/
https://codereview.chromium.org/207153003/

[Cutch, 2014-03-21 18:08:51]

lgtm and thanks!

[koda, 2014-03-21 21:12:41]

Committed patchset #2 manually as r34271 (presubmit successful).

[turnidge, 2014-03-23 22:24:19]

I know you've committed this, but...

https://codereview.chromium.org/206003005/diff/20001/runtime/vm/object.cc
File runtime/vm/object.cc (right):

https://codereview.chromium.org/206003005/diff/20001/runtime/vm/object.cc#new...
runtime/vm/object.cc:6328: }
Did you consider grabbing an id from the object id ring but not delegating up to
the Object::PrintToJSONStream.  Something like:

if (id >= 0) {
  jsobj.AddPropertyF("id", "classes/%" Pd " ... etc.
} else {
  jsobj.AddPropertyF("id", "objects/%" Pd "", <get a number from the obj id
ring>);
}

That way this will have the expected format of a Function and only the id will
be different.  Otherwise we will have to guard against these odd functions in
the observatory.

https://codereview.chromium.org/206003005/diff/20001/runtime/vm/object.cc#new...
runtime/vm/object.cc:13492: }
Ditto previous comment.  You could print an id from the object id ring but print
the rest of the expected properties for a Type.

[koda, 2014-03-24 17:18:34]

https://codereview.chromium.org/206003005/diff/20001/runtime/vm/object.cc
File runtime/vm/object.cc (right):

https://codereview.chromium.org/206003005/diff/20001/runtime/vm/object.cc#new...
runtime/vm/object.cc:6328: }
On 2014/03/23 22:24:20, turnidge wrote:
> Did you consider grabbing an id from the object id ring but not delegating up
to
> the Object::PrintToJSONStream.  Something like:
> 
> if (id >= 0) {
>   jsobj.AddPropertyF("id", "classes/%" Pd " ... etc.
> } else {
>   jsobj.AddPropertyF("id", "objects/%" Pd "", <get a number from the obj id
> ring>);
> }
> 
> That way this will have the expected format of a Function and only the id will
> be different.  Otherwise we will have to guard against these odd functions in
> the observatory.

I wasn't sure whether all fields still work in all the special cases (e.g., will
things like "parent_function" work?), but I suppose it makes sense to try.

I'll send a follow-up CL.

https://codereview.chromium.org/206003005/diff/20001/runtime/vm/object.cc#new...
runtime/vm/object.cc:13492: }
On 2014/03/23 22:24:20, turnidge wrote:
> Ditto previous comment.  You could print an id from the object id ring but
print
> the rest of the expected properties for a Type.

Also, here: not sure whether all printed properties are even defined for a
non-finalized type, for instance.