Fixing up measurement test.

Fixing up measurement test.

The test was trying to validate that postMessage is called back before setTimeout(0), but there is no guarantee of this by the browsers and the behavior is inconsistent (even within the context of a single browser).


BUG=1940
Committed: https://code.google.com/p/dart/source/detail?r=12723

[Emily Fortuna, 2012-09-18 21:19:30]

lgtm. Thanks for tracking this down!

[vsm, 2012-09-19 15:43:59]

[+jacobr who i think wrote the original test]

Jacob: I think I asked you this before: is there a reason you wanted this to
fire before setTimeout(0)?

[blois, 2012-09-19 16:28:10]

It seems like it should be fired before setTimeout(0), but the underlying
mechanism uses postMessage and I can find nothing which states that it must fire
before setTimeout.

I put together a simple test which does a setTimeout and a postMessage and
checks the order which they are called, then repeats. For the most part,
postMessage is called before setTimeout, but by no means all the time. It seems
that for a short while after startup Chrome will process setTimeout first, and
Firefox occasionally does setTimeout first through the life of the app.

On 2012/09/19 15:43:59, vsm wrote:
> [+jacobr who i think wrote the original test]
> 
> Jacob: I think I asked you this before: is there a reason you wanted this to
> fire before setTimeout(0)?

[blois, 2012-09-20 16:59:14]

Changing test back closer to original form, but using MutationObservers to
generate the callbacks, per Jacob's suggestion.

On 2012/09/19 16:28:10, blois wrote:
> It seems like it should be fired before setTimeout(0), but the underlying
> mechanism uses postMessage and I can find nothing which states that it must
fire
> before setTimeout.
> 
> I put together a simple test which does a setTimeout and a postMessage and
> checks the order which they are called, then repeats. For the most part,
> postMessage is called before setTimeout, but by no means all the time. It
seems
> that for a short while after startup Chrome will process setTimeout first, and
> Firefox occasionally does setTimeout first through the life of the app.
> 
> On 2012/09/19 15:43:59, vsm wrote:
> > [+jacobr who i think wrote the original test]
> > 
> > Jacob: I think I asked you this before: is there a reason you wanted this to
> > fire before setTimeout(0)?

[vsm, 2012-09-21 15:12:40]

lgtm!

https://chromiumcodereview.appspot.com/10951016/diff/8001/lib/html/src/Measur...
File lib/html/src/Measurement.dart (right):

https://chromiumcodereview.appspot.com/10951016/diff/8001/lib/html/src/Measur...
lib/html/src/Measurement.dart:86: this._onCallback();
Should this check for a "DART-MEASURE" message?  It actually doesn't seem to
matter.

https://chromiumcodereview.appspot.com/10951016/diff/8001/lib/html/src/dart2j...
File lib/html/src/dart2js_MutationObserverSupported.dart (right):

https://chromiumcodereview.appspot.com/10951016/diff/8001/lib/html/src/dart2j...
lib/html/src/dart2js_MutationObserverSupported.dart:9: static bool
_isMutationObserverSupported() native '''
We should think of a better and user-accessible pattern for this type of check.

[blois, 2012-09-21 17:00:03]

https://codereview.chromium.org/10951016/diff/8001/lib/html/src/Measurement.dart
File lib/html/src/Measurement.dart (right):

https://codereview.chromium.org/10951016/diff/8001/lib/html/src/Measurement.d...
lib/html/src/Measurement.dart:86: this._onCallback();
On 2012/09/21 15:12:40, vsm wrote:
> Should this check for a "DART-MEASURE" message?  It actually doesn't seem to
> matter.

Nope, there's a comment where it sends the message- since this is just looking
for the next message, it triggers off of any message.

https://codereview.chromium.org/10951016/diff/8001/lib/html/src/dart2js_Mutat...
File lib/html/src/dart2js_MutationObserverSupported.dart (right):

https://codereview.chromium.org/10951016/diff/8001/lib/html/src/dart2js_Mutat...
lib/html/src/dart2js_MutationObserverSupported.dart:9: static bool
_isMutationObserverSupported() native '''
On 2012/09/21 15:12:40, vsm wrote:
> We should think of a better and user-accessible pattern for this type of
check.

Agree- generalizing a mechanism for checking whether features are supported is
for a later milestone.