fixes dart.throw to work for non-objects

fixes dart.throw to work for non-objects
language tests use this, for example

R=vsm@google.com

Committed: https://github.com/dart-lang/dev_compiler/commit/eecdbe4fed26f37e020eb2315f532794cfd35588

[Jennifer Messerly, 2015-07-22 23:19:40]

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

[Jennifer Messerly, 2015-07-22 23:21:30]

on weakmap: it should be a bit better for expando like things added to object
instances. it won't cause the object to need a new class. (the other places we
add props, we do it to the prototype at set up time)

[vsm, 2015-07-23 14:00:04]

lgtm

https://codereview.chromium.org/1245023004/diff/20001/lib/runtime/_operations.js
File lib/runtime/_operations.js (right):

https://codereview.chromium.org/1245023004/diff/20001/lib/runtime/_operations...
lib/runtime/_operations.js:277: // allow it to work for things like strings and
numbers.
I don't think we want this in the general case.  I encounter a lot of wrapped
exceptions (e.g., in async code), where it's really useful to look at the inner
exception's trace.

Another possibility is to box the string, etc., if we encounter a primitive:

obj = Object(obj);

and throw that instead.

[Jennifer Messerly, 2015-07-23 15:51:30]

https://codereview.chromium.org/1245023004/diff/20001/lib/runtime/_operations.js
File lib/runtime/_operations.js (right):

https://codereview.chromium.org/1245023004/diff/20001/lib/runtime/_operations...
lib/runtime/_operations.js:277: // allow it to work for things like strings and
numbers.
On 2015/07/23 14:00:04, vsm wrote:
> I don't think we want this in the general case.  I encounter a lot of wrapped
> exceptions (e.g., in async code), where it's really useful to look at the
inner
> exception's trace.

Sure, but they should be wrapping the stack trace parameter too? Then they'll
have saved the original inner stack trace, and it will all work.

> 
> Another possibility is to box the string, etc., if we encounter a primitive:
> 
> obj = Object(obj);
> 
> and throw that instead.

eeep. I suspect that would break a lot of things

[Jennifer Messerly, 2015-07-23 16:34:45]

Committed patchset #2 (id:20001) manually as
eecdbe4fed26f37e020eb2315f532794cfd35588 (presubmit successful).