We have to create different locals when analyzing a closure. Otherwise the inferrer visitor might t…

We have to create different locals when analyzing a closure. Otherwise the inferrer visitor might think an aborting body in a closure aborts the enclosing function.

R=karlklose@google.com

Committed: https://code.google.com/p/dart/source/detail?r=24181

[karlklose, 2013-06-19 13:13:13]

LGTM.

https://codereview.chromium.org/16959009/diff/1/sdk/lib/_internal/compiler/im...
File sdk/lib/_internal/compiler/implementation/types/container_tracer.dart
(right):

https://codereview.chromium.org/16959009/diff/1/sdk/lib/_internal/compiler/im...
sdk/lib/_internal/compiler/implementation/types/container_tracer.dart:345: new
ContainerTracerVisitor(function, tracer, closureLocals).run();
Would pushing the locals on a stack work instead of keep pushing visitors on the
callstack?

[ngeoffray, 2013-06-19 13:20:01]

Thanks Karl!

https://codereview.chromium.org/16959009/diff/1/sdk/lib/_internal/compiler/im...
File sdk/lib/_internal/compiler/implementation/types/container_tracer.dart
(right):

https://codereview.chromium.org/16959009/diff/1/sdk/lib/_internal/compiler/im...
sdk/lib/_internal/compiler/implementation/types/container_tracer.dart:345: new
ContainerTracerVisitor(function, tracer, closureLocals).run();
On 2013/06/19 13:13:13, karlklose wrote:
> Would pushing the locals on a stack work instead of keep pushing visitors on
the
> callstack?

Yes, I believe it would. I went that way because that's how it's also done in
the inferrer. Also I find it less error-prone when you e.g. directly know when
visiting a return that you're returning from the current method of the visitor.

[ngeoffray, 2013-06-19 13:21:14]

Committed patchset #2 manually as r24181 (presubmit successful).