Make VisitStatements() consistent among all AstVisitor implementations

Make VisitStatements() consistent among all AstVisitor implementations

R=yangguo@chromium.org

Committed: http://code.google.com/p/v8/source/detail?r=16443

[Jakob Kummerow, 2013-08-30 10:34:48]

AstTyper::VisitStatements() contains the optimization "if (stmt->IsJump())
break;". Other AstVisitors, in particular the HOptimizedGraphBuilder, must do
the same, or they will stumble over uninitialized type feedback (empty handles)
in nodes that the typer has skipped.

An alternative implementation would be to remove the early return from the
AstTyper, but I think it's a valid optimization, so we should just do it
everywhere.

[Yang, 2013-08-30 10:37:48]

On 2013/08/30 10:34:48, Jakob wrote:
> AstTyper::VisitStatements() contains the optimization "if (stmt->IsJump())
> break;". Other AstVisitors, in particular the HOptimizedGraphBuilder, must do
> the same, or they will stumble over uninitialized type feedback (empty
handles)
> in nodes that the typer has skipped.
> 
> An alternative implementation would be to remove the early return from the
> AstTyper, but I think it's a valid optimization, so we should just do it
> everywhere.

lgtm.

[Jakob Kummerow, 2013-08-30 10:51:47]

Committed patchset #1 manually as r16443.

[rossberg, 2013-08-30 11:24:21]

On 2013/08/30 10:34:48, Jakob wrote:
> AstTyper::VisitStatements() contains the optimization "if (stmt->IsJump())
> break;". Other AstVisitors, in particular the HOptimizedGraphBuilder, must do
> the same, or they will stumble over uninitialized type feedback (empty
handles)
> in nodes that the typer has skipped.
> 
> An alternative implementation would be to remove the early return from the
> AstTyper, but I think it's a valid optimization, so we should just do it
> everywhere.

LGTM too. Thanks for fixing.

Ideally, we should really only have one implementation of
VisitStatements/Declarations and the CHECK macros that works for all visitors.
Maybe I'll try to refactor that.