Restart AST node numbering when we enter a function.

Restart AST node numbering when we enter a function.


BUG=
TEST=

Committed: http://code.google.com/p/v8/source/detail?r=7543

[William Hesse, 2011-04-05 15:23:57]

Not ready for commit yet.  Some failures, and I want to create a scoped object
to reset the AST Id to 1 and save the old one.  Can the scope be narrowed to
just the one surrounding the statement list, or are the parameters and local
variables being given AST Ids?

[Kevin Millikin (Chromium), 2011-04-05 15:56:11]

http://codereview.chromium.org/6691058/diff/1/src/parser.cc
File src/parser.cc (right):

http://codereview.chromium.org/6691058/diff/1/src/parser.cc#newcode3548
src/parser.cc:3548: Isolate* isolate = Isolate::Current();
You should probably call (the parser's) isolate() function here to use the
cached one.

I think you should move all this code into the constructor for LexicalScope. 
You probably have to shorten its scope a tiny bit and read out one field before
it goes out of scope (ContainsLoops()).

(Actually, I don't think FunctionLiteral::contains_loops() is ever called,
either.  Delete it.)

[William Hesse, 2011-04-07 12:27:09]

All comments addressed, moved into LexicalScope, and loop counting removed from
parser and lexical scope.

[Kevin Millikin (Chromium), 2011-04-07 13:42:51]

LGTM.  Suggestion to change the name of the constant.

http://codereview.chromium.org/6691058/diff/8005/src/arm/full-codegen-arm.cc
File src/arm/full-codegen-arm.cc (right):

http://codereview.chromium.org/6691058/diff/8005/src/arm/full-codegen-arm.cc#...
src/arm/full-codegen-arm.cc:248: PrepareForBailoutForId(AstNode::kRootAstId,
NO_REGISTERS);
How about something more like kFunctionEntryId?

http://codereview.chromium.org/6691058/diff/8005/src/ast.h
File src/ast.h (right):

http://codereview.chromium.org/6691058/diff/8005/src/ast.h#newcode137
src/ast.h:137: // The root of every AST, a FunctionLiteral node, has id
kRootAstId when
This comment is unnecessary and I think it confuses matters.  The function
literal does not have two different id_ values depending on the observer.  It
has one.  We don't (shouldn't) use it in the function itself.

There is a separate function entry id.  Call it something like kFunctionEntryId
and you probably don't even need a comment it's so obvious.

The fact that we used to use the literal's id because it wasn't used for
anything else is not important.

http://codereview.chromium.org/6691058/diff/8005/src/parser.cc
File src/parser.cc (right):

http://codereview.chromium.org/6691058/diff/8005/src/parser.cc#newcode3683
src/parser.cc:3683: FunctionLiteral* function_literal = new(zone())
FunctionLiteral(
I sort of prefer breaking at the lowest precedence operator, if it makes the
lines fit and I don't like the offsides look of the (.  In this case, it should
work to have:

FunctionLiteral* function_literal =
  new(zone()) FunctionLiteral(name,
                              top_scope_,
...