Support for native in leg, and start moving native tests to a specific test suite.

Support for native in leg, and start moving native tests to a specific test suite.
Committed: https://code.google.com/p/dart/source/detail?r=4392

[floitsch, 2012-02-19 00:59:43]

DBC.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dart
File frog/leg/emitter.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:27: if (typeof(constructor) == 'function') {
typeof is not a function:
if (typeof constructor == 'function')

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:36: if (name && typeof(name) == 'string' && name !=
'Object') return name;
exchange typeof and name-check.
It probably does not matter, but it feels righter to me :)

(background: typeof check is cheap, and the boolification of the string should
be cheaper then).

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:55: String get DYNAMIC_FUNCTION() => '''
this needs comments.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:309: if (codeBlock !== null) {
Maybe we should consider making our bailout function static.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:351: return generateNativeClass(classElement, buffer);
don't use 'return val' to leave a void function.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/lis...
File frog/leg/scanner/listener.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/lis...
frog/leg/scanner/listener.dart:436: error("expected a block '$token'", token);
but got '$token'

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/lis...
frog/leg/scanner/listener.dart:441: error("expected a function body '$token'",
token);
but got

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/lis...
frog/leg/scanner/listener.dart:446: error("expected a class body '$token'",
token);
but got.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/ssa/builder...
File frog/leg/ssa/builder.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/ssa/builder...
frog/leg/ssa/builder.dart:1573: List<String> buffer = <String>[];
Don't call a list a buffer.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/ssa/builder...
frog/leg/ssa/builder.dart:1588: 'return
$receiver${work.element.name}($foreignParameters)');
HForeign should be an expression.
There should be no 'return' here. If necessary add an HReturn.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/ssa/codegen...
File frog/leg/ssa/codegen.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/ssa/codegen...
frog/leg/ssa/codegen.dart:539: buffer.add('$code');
Afair we do 'generateAtUseSite' for HForeigns. Unless you can guarantee that
there will never be a precedence issue I vote strongly in favor of keeping the
parenthesis.
As said before HForeign should be expressions.

[ahe, 2012-02-19 14:38:02]

I'm not sure about this change because:

* I understand the purpose of DYNAMIC_FUNCTION (lack of documentation).

* I'm not convinced that your changes are being tested.

Otherwise, this look very nice.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/compiler.dart
File frog/leg/compiler.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/compiler.da...
frog/leg/compiler.dart:256: if
(script.name.contains('dart/frog/tests/native/src')) {
I like this.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dart
File frog/leg/emitter.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:25: function(obj) {
Could you add a documentation comment for this function?

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:26: var constructor = obj.constructor;
What happens if obj is undefined?

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:39: var name = string.substring(8, string.length - 1);
I'm getting a feeling that this is very similar to Primitives.objectToString in
js_helper. We should consolidate eventually.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:55: String get DYNAMIC_FUNCTION() => '''
Documentation comment, please.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:89: if (typeof $dynamicMetadataName == 'undefined')
$dynamicMetadataName = [];
Extra code after function. Shouldn't this be in the isolate?

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:99: String get inheritsName() =>
'${compiler.namer.ISOLATE}.\$inherits';
Why aren't these strings final values? Generating them repeatedly seems
wasteful. I guess the idea is that they are only called once, however, the
pattern is confusing.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:104: '${compiler.namer.ISOLATE}.\$dynamicMetatada';
How is this related to line 89?

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:143: buffer.add('${attachTo(invocationName)} =
function(');
How about turning this into a streaming API, that is, something like this:

attachNameOn(buffer, invocationName);

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:315: if (!parameters.optionalParameters.isEmpty()) {
Is this necessary if there is no code generated?

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/lis...
File frog/leg/scanner/listener.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/lis...
frog/leg/scanner/listener.dart:529: String nativeName;
SourceString.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/lis...
frog/leg/scanner/listener.dart:533: nativeName =
nodes.head.asIdentifier().token.next.stringValue;
Don't call stringValue.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/lis...
frog/leg/scanner/listener.dart:695: if (token.stringValue === 'native') {
Should we move this code to some static helpers in a different file? Then we can
keep this native stuff more isolated.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/lis...
frog/leg/scanner/listener.dart:701: BeginGroupToken beginGroupToken = token;
Why is the block being skipped here? 

Perhaps this code belongs somewhere else.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/lis...
frog/leg/scanner/listener.dart:726: BeginGroupToken beginGroupToken = token;
Ditto.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/par...
File frog/leg/scanner/parser.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/par...
frog/leg/scanner/parser.dart:200: return listener.expectedBlock(token);
I think this event should be "expectedBlockToSkip" or something similar.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/par...
frog/leg/scanner/parser.dart:215: Token id = token.next;
I don't think id is used.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/par...
File frog/leg/scanner/partial_parser.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/par...
frog/leg/scanner/partial_parser.dart:35: return
listener.expectedClassBody(token);
Event should be something like "expectedClassBodyToSkip" or similar.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/sca...
File frog/leg/scanner/scanner_task.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/sca...
frog/leg/scanner/scanner_task.dart:15: for (Link<Element> link =
compilationUnit.topLevelElements;
This should be in the if case above. Also, I'd like to have something like:

processNativeClasses(compilationUnit);

Where "processNativeClasses" is in a separate file.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/presubmit.py
File frog/presubmit.py (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/presubmit.py#ne...
frog/presubmit.py:148: 'isolate', 'peg', 'frog', 'css', 'dartdoc', 'native']
What about leg?

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/tests/native/na...
File frog/tests/native/native.status (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/tests/native/na...
frog/tests/native/native.status:5: prefix native
I think this should be called "frog_native". This is not a general "native"
mechanism.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/tests/native/te...
File frog/tests/native/test_config.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/tests/native/te...
frog/tests/native/test_config.dart:9: class FrogNativeTestSuite extends
StandardTestSuite {
I think you need to change something .../tools/test.dart as well. I don't think
you're running the tests right now.

[ahe, 2012-02-19 14:38:33]

On 2012/02/19 14:38:02, ahe wrote:
> * I understand the purpose of DYNAMIC_FUNCTION (lack of documentation).

I meant: I *don't* understand the purpose of DYNAMIC_FUNCTION (lack of
documentation).

[Lasse Reichstein Nielsen, 2012-02-20 07:33:56]

dbc.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/compiler.dart
File frog/leg/compiler.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/compiler.da...
frog/leg/compiler.dart:256: if
(script.name.contains('dart/frog/tests/native/src')) {
I would prefer to have it only enabled by a flag (possibly even have the flag
contain the path slice instead of hardcoding it).
It's a magic value, and a test that can potentially blow up in someone's face.

[ngeoffray, 2012-02-20 09:09:37]

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dart
File frog/leg/emitter.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:25: function(obj) {
On 2012/02/19 14:38:02, ahe wrote:
> Could you add a documentation comment for this function?

Sure. For the reference, this code is a pure copy and paste of frog's code.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:26: var constructor = obj.constructor;
On 2012/02/19 14:38:02, ahe wrote:
> What happens if obj is undefined?

boom? :)

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:27: if (typeof(constructor) == 'function') {
On 2012/02/19 00:59:43, floitsch wrote:
> typeof is not a function:
> if (typeof constructor == 'function')

Copy/pasted code.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:36: if (name && typeof(name) == 'string' && name !=
'Object') return name;
On 2012/02/19 00:59:43, floitsch wrote:
> exchange typeof and name-check.
> It probably does not matter, but it feels righter to me :)
> 
> (background: typeof check is cheap, and the boolification of the string should
> be cheaper then).

Copy-pasted code.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:39: var name = string.substring(8, string.length - 1);
On 2012/02/19 14:38:02, ahe wrote:
> I'm getting a feeling that this is very similar to Primitives.objectToString
in
> js_helper. We should consolidate eventually.

Not on this code, which is copy-pasted.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:55: String get DYNAMIC_FUNCTION() => '''
On 2012/02/19 00:59:43, floitsch wrote:
> this needs comments.

As for TYPE_NAME_OF_FUNCTION, this method was taken from frog.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:89: if (typeof $dynamicMetadataName == 'undefined')
$dynamicMetadataName = [];
On 2012/02/19 14:38:02, ahe wrote:
> Extra code after function. Shouldn't this be in the isolate?

Code also comes from frog.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:99: String get inheritsName() =>
'${compiler.namer.ISOLATE}.\$inherits';
On 2012/02/19 14:38:02, ahe wrote:
> Why aren't these strings final values? Generating them repeatedly seems
> wasteful. I guess the idea is that they are only called once, however, the
> pattern is confusing.

They cannot be final because compiler.namer.ISOLATE is not final.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:104: '${compiler.namer.ISOLATE}.\$dynamicMetatada';
On 2012/02/19 14:38:02, ahe wrote:
> How is this related to line 89?

Not sure I understand.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:315: if (!parameters.optionalParameters.isEmpty()) {
On 2012/02/19 14:38:02, ahe wrote:
> Is this necessary if there is no code generated?

Very good catch. Will remove it.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/lis...
File frog/leg/scanner/listener.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/lis...
frog/leg/scanner/listener.dart:701: BeginGroupToken beginGroupToken = token;
On 2012/02/19 14:38:02, ahe wrote:
> Why is the block being skipped here? 

The block contains native code, and I need to return the end of the block to
continue parsing.
> 
> Perhaps this code belongs somewhere else.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/lis...
frog/leg/scanner/listener.dart:726: BeginGroupToken beginGroupToken = token;
On 2012/02/19 14:38:02, ahe wrote:
> Ditto.

Ditto, I need to return the end of the class body to continue diet parsing.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/sca...
File frog/leg/scanner/scanner_task.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/sca...
frog/leg/scanner/scanner_task.dart:15: for (Link<Element> link =
compilationUnit.topLevelElements;
On 2012/02/19 14:38:02, ahe wrote:
> This should be in the if case above. Also, I'd like to have something like:

How about elements defined in a non-library compilation unit?

> 
> processNativeClasses(compilationUnit);
> 
> Where "processNativeClasses" is in a separate file.

Will do.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/ssa/builder...
File frog/leg/ssa/builder.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/ssa/builder...
frog/leg/ssa/builder.dart:1588: 'return
$receiver${work.element.name}($foreignParameters)');
On 2012/02/19 00:59:43, floitsch wrote:
> HForeign should be an expression.
> There should be no 'return' here. If necessary add an HReturn.

As discussed, the code in the native is not just an expression, it could be a
full body of Js code. Will discuss on how to proceed.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/presubmit.py
File frog/presubmit.py (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/presubmit.py#ne...
frog/presubmit.py:148: 'isolate', 'peg', 'frog', 'css', 'dartdoc', 'native']
On 2012/02/19 14:38:02, ahe wrote:
> What about leg?

Ah good catch! Did not realize this was only if non leg-only :)

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/tests/native/te...
File frog/tests/native/test_config.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/tests/native/te...
frog/tests/native/test_config.dart:9: class FrogNativeTestSuite extends
StandardTestSuite {
On 2012/02/19 14:38:02, ahe wrote:
> I think you need to change something .../tools/test.dart as well. I don't
think
> you're running the tests right now.

Ah, I did change them, but did a gcl change in the frog directory that did not
include the top-level files. So the tests were run (thanks to your catch, not in
presubmit though).

[ahe, 2012-02-20 09:36:10]

Sorry, I still don't understand the code.

Also, I think you also need to modify
dart/frog/scripts/buildbot_annotated_steps.py to run the tests on leg.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dart
File frog/leg/emitter.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:55: String get DYNAMIC_FUNCTION() => '''
On 2012/02/20 09:09:37, ngeoffray wrote:
> On 2012/02/19 00:59:43, floitsch wrote:
> > this needs comments.
> 
> As for TYPE_NAME_OF_FUNCTION, this method was taken from frog.

Whatever information you have about these functions, please document it. Perhaps
you can ask Stephen for more information.

You must have at least some idea about what is going on here.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:89: if (typeof $dynamicMetadataName == 'undefined')
$dynamicMetadataName = [];
On 2012/02/20 09:09:37, ngeoffray wrote:
> On 2012/02/19 14:38:02, ahe wrote:
> > Extra code after function. Shouldn't this be in the isolate?
> 
> Code also comes from frog.

You chose the word "DYNAMIC_FUNCTION". Clearly this is not just a function.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:104: '${compiler.namer.ISOLATE}.\$dynamicMetatada';
On 2012/02/20 09:09:37, ngeoffray wrote:
> On 2012/02/19 14:38:02, ahe wrote:
> > How is this related to line 89?
> 
> Not sure I understand.

I overlooked that the code on 89 wasn't in a raw string.

I think this coding style is confusing, and it would be clearer if
DYNAMIC_FUNCTION didn't pretend to be a constant and instead was a real method
with parameters.

[ahe, 2012-02-20 10:02:55]

Since I'm not in the office today and we can't talk about this in person, I'm
saying LGTM provided you can find someone else who is a little more comfortable
with the dynamic function stuff.

[ngeoffray, 2012-02-20 17:00:01]

PTAL

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/compiler.dart
File frog/leg/compiler.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/compiler.da...
frog/leg/compiler.dart:256: if
(script.name.contains('dart/frog/tests/native/src')) {
On 2012/02/20 07:33:56, Lasse Reichstein Nielsen wrote:
> I would prefer to have it only enabled by a flag (possibly even have the flag
> contain the path slice instead of hardcoding it).
> It's a magic value, and a test that can potentially blow up in someone's face.

As discussed, that's half-way intended :)

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/lis...
File frog/leg/scanner/listener.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/lis...
frog/leg/scanner/listener.dart:436: error("expected a block '$token'", token);
On 2012/02/19 00:59:43, floitsch wrote:
> but got '$token'

Done.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/lis...
frog/leg/scanner/listener.dart:441: error("expected a function body '$token'",
token);
On 2012/02/19 00:59:43, floitsch wrote:
> but got

Done.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/lis...
frog/leg/scanner/listener.dart:446: error("expected a class body '$token'",
token);
On 2012/02/19 00:59:43, floitsch wrote:
> but got.

Done.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/lis...
frog/leg/scanner/listener.dart:529: String nativeName;
On 2012/02/19 14:38:02, ahe wrote:
> SourceString.

Done.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/lis...
frog/leg/scanner/listener.dart:533: nativeName =
nodes.head.asIdentifier().token.next.stringValue;
On 2012/02/19 14:38:02, ahe wrote:
> Don't call stringValue.

Done.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/lis...
frog/leg/scanner/listener.dart:695: if (token.stringValue === 'native') {
On 2012/02/19 14:38:02, ahe wrote:
> Should we move this code to some static helpers in a different file? Then we
can
> keep this native stuff more isolated.

Done.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/par...
File frog/leg/scanner/parser.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/par...
frog/leg/scanner/parser.dart:200: return listener.expectedBlock(token);
On 2012/02/19 14:38:02, ahe wrote:
> I think this event should be "expectedBlockToSkip" or something similar.

Done.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/par...
frog/leg/scanner/parser.dart:215: Token id = token.next;
On 2012/02/19 14:38:02, ahe wrote:
> I don't think id is used.

Done.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/par...
File frog/leg/scanner/partial_parser.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/scanner/par...
frog/leg/scanner/partial_parser.dart:35: return
listener.expectedClassBody(token);
On 2012/02/19 14:38:02, ahe wrote:
> Event should be something like "expectedClassBodyToSkip" or similar.

Done.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/ssa/builder...
File frog/leg/ssa/builder.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/ssa/builder...
frog/leg/ssa/builder.dart:1573: List<String> buffer = <String>[];
On 2012/02/19 00:59:43, floitsch wrote:
> Don't call a list a buffer.

Done.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/ssa/builder...
frog/leg/ssa/builder.dart:1588: 'return
$receiver${work.element.name}($foreignParameters)');
On 2012/02/19 00:59:43, floitsch wrote:
> HForeign should be an expression.
> There should be no 'return' here. If necessary add an HReturn.

Done.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/ssa/codegen...
File frog/leg/ssa/codegen.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/ssa/codegen...
frog/leg/ssa/codegen.dart:539: buffer.add('$code');
On 2012/02/19 00:59:43, floitsch wrote:
> Afair we do 'generateAtUseSite' for HForeigns. Unless you can guarantee that
> there will never be a precedence issue I vote strongly in favor of keeping the
> parenthesis.
> As said before HForeign should be expressions.

Added a test generateAtUseSite to know if we need to add parens.

[floitsch, 2012-02-21 13:11:15]

LGTM.

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dart
File frog/leg/emitter.dart (right):

https://chromiumcodereview.appspot.com/9418045/diff/2001/frog/leg/emitter.dar...
frog/leg/emitter.dart:26: var constructor = obj.constructor;
On 2012/02/20 09:09:37, ngeoffray wrote:
> On 2012/02/19 14:38:02, ahe wrote:
> > What happens if obj is undefined?
> 
> boom? :)

Note that this function is called on the prototype of Object and 'obj' is set to
'this' there.
So null is not possible.

[ahe, 2012-02-22 22:08:47]

https://chromiumcodereview.appspot.com/9418045/diff/11002/frog/scripts/buildb...
File frog/scripts/buildbot_annotated_steps.py (right):

https://chromiumcodereview.appspot.com/9418045/diff/11002/frog/scripts/buildb...
frog/scripts/buildbot_annotated_steps.py:156: TestStep("leg_extra", mode,
system, 'leg', ['leg_only'], flags)
You need to pass in 'frog_native' here. Otherwise the tests won't get run on the
build bot.