Re-try of Tightening up dart2js load timing.

Re-try of Tightening up dart2js load timing.

This is a fixup of https://codereview.chromium.org/22488002/, first patch is the original, subsequent the fixes.

BUG=
R=ahe@google.com, vsm@google.com

Committed: https://code.google.com/p/dart/source/detail?r=26244

[blois, 2013-08-15 21:38:48]

I have not gotten my enlistment working on IE yet, so not sure if this is the
issue which is causing it to fail there, but this does cause the timeouts from
FF when executed via the browser_controller:

./tools/test.py -c dart2js -r ff --mode=release --checked --arch=ia32
--use_browser_controller -v -t60000 -j1 html/url_test

https://codereview.chromium.org/23121004/diff/16001/pkg/browser/lib/dart.js
File pkg/browser/lib/dart.js (right):

https://codereview.chromium.org/23121004/diff/16001/pkg/browser/lib/dart.js#n...
pkg/browser/lib/dart.js:30: var parent = scripts[i].parentNode;
This parent variable was leaking to the window, overriding window.parent.

This then broke test_controller.js, which has:
function getDriverWindow() {
  if (window != window.parent) {
    // We're running in an iframe.
    return window.parent;
  } else if (window.opener) {
    // We were opened by another window.
    return window.opener;
  }
  return null;
}

window.parent was 'undefined', and so this function returned undefined and test
messages were not getting sent.

[vsm, 2013-08-15 22:05:59]

lgtm

[ahe, 2013-08-15 22:20:27]

LGTM!

https://codereview.chromium.org/23121004/diff/16001/sdk/lib/_internal/compile...
File sdk/lib/_internal/compiler/implementation/js_backend/emitter.dart (right):

https://codereview.chromium.org/23121004/diff/16001/sdk/lib/_internal/compile...
sdk/lib/_internal/compiler/implementation/js_backend/emitter.dart:2982:
;(function (callback) {
Why the extra semicolon?

https://codereview.chromium.org/23121004/diff/16001/sdk/lib/_internal/compile...
sdk/lib/_internal/compiler/implementation/js_backend/emitter.dart:3017: });
Use $N instead of semicolon-newline here.

[ahe, 2013-08-15 22:24:38]

https://codereview.chromium.org/23121004/diff/16001/sdk/lib/_internal/compile...
File sdk/lib/_internal/compiler/implementation/js_backend/emitter.dart (right):

https://codereview.chromium.org/23121004/diff/16001/sdk/lib/_internal/compile...
sdk/lib/_internal/compiler/implementation/js_backend/emitter.dart:2993: function
onLoad() {
Shouldn't this function have a parameter named 'event'?

[blois, 2013-08-15 22:38:48]

https://codereview.chromium.org/23121004/diff/16001/sdk/lib/_internal/compile...
File sdk/lib/_internal/compiler/implementation/js_backend/emitter.dart (right):

https://codereview.chromium.org/23121004/diff/16001/sdk/lib/_internal/compile...
sdk/lib/_internal/compiler/implementation/js_backend/emitter.dart:2982:
;(function (callback) {
On 2013/08/15 22:20:27, ahe wrote:
> Why the extra semicolon?

In minified mode, the preceeding lines look like:
init.metadata=[];$=null
I = I.$finishIsolateConstructor(I)
$=new I()

The closure for this gets interpreted as a function call on the I() rather than
a new function.

https://codereview.chromium.org/23121004/diff/16001/sdk/lib/_internal/compile...
sdk/lib/_internal/compiler/implementation/js_backend/emitter.dart:2993: function
onLoad() {
On 2013/08/15 22:24:38, ahe wrote:
> Shouldn't this function have a parameter named 'event'?

Great catch- that must be the IE issue.

https://codereview.chromium.org/23121004/diff/16001/sdk/lib/_internal/compile...
sdk/lib/_internal/compiler/implementation/js_backend/emitter.dart:3017: });
On 2013/08/15 22:20:27, ahe wrote:
> Use $N instead of semicolon-newline here.

Done.

[ahe, 2013-08-15 22:43:48]

LGTM

https://codereview.chromium.org/23121004/diff/16001/sdk/lib/_internal/compile...
File sdk/lib/_internal/compiler/implementation/js_backend/emitter.dart (right):

https://codereview.chromium.org/23121004/diff/16001/sdk/lib/_internal/compile...
sdk/lib/_internal/compiler/implementation/js_backend/emitter.dart:2982:
;(function (callback) {
On 2013/08/15 22:38:48, blois wrote:
> On 2013/08/15 22:20:27, ahe wrote:
> > Why the extra semicolon?
> 
> In minified mode, the preceeding lines look like:
> init.metadata=[];$=null
> I = I.$finishIsolateConstructor(I)
> $=new I()
> 
> The closure for this gets interpreted as a function call on the I() rather
than
> a new function.

I think you should modify line 3865 instead. It should be ;$n instead of $N.

[blois, 2013-08-15 23:57:16]

Committed patchset #4 manually as r26244 (presubmit successful).

[ricow1, 2013-08-16 05:33:55]

Thanks a lot for fixing this Pete