Fixing the detection of aliased eval so that it is exact....

Fixing the detection of aliased eval so that it is exact.
Fixing the semantics of aliased eval so that it is conformant.

Committed: http://code.google.com/p/v8/source/detail?r=819

[Mads Ager (chromium), 2008-11-21 10:53:23]

This is a good step in the right direction.  Two main issues: 

1. global variables should be thread local

2. the IsUnaliasedEval runtime routine needs to check whether eval was found not
only in a global object, but also that the global object was at the end of the
scope chain.

http://codereview.chromium.org/11563/diff/1/4
File src/ast.h (right):

http://codereview.chromium.org/11563/diff/1/4#newcode869
Line 869: GLOBAL_EVAL     // An eval that statically refers to the unaliased,
Is this comment correct? We still need a check at runtime, right?  So this type
is for potential direct eval calls.  Maybe the enum value name could reflect
that too.

Would it be clearer to call this EvalType with the values ALIASED and
MAYBE_DIRECT?

http://codereview.chromium.org/11563/diff/1/8
File src/codegen-arm.cc (right):

http://codereview.chromium.org/11563/diff/1/8#newcode2324
Line 2324: __ CallRuntime(Runtime::kSetInGlobalEval, 0);
Could we call this kSetPotentialDirectEvalCall or something like that? 
Similarly for the clear?

http://codereview.chromium.org/11563/diff/1/10
File src/parser.cc (right):

http://codereview.chromium.org/11563/diff/1/10#newcode329
Line 329: 
Remove the extra empty line here?

http://codereview.chromium.org/11563/diff/1/10#newcode2303
Line 2303: // the current scope chain. These calls are marked as global eval
potential direct eval call?

http://codereview.chromium.org/11563/diff/1/5
File src/runtime.cc (right):

http://codereview.chromium.org/11563/diff/1/5#newcode3992
Line 3992: static bool inGlobalEval = false;
This global variable needs to be a part of the thread local state in top.[h,cc]
so that multiple threads using V8 one at a time will not influence each other.

http://codereview.chromium.org/11563/diff/1/5#newcode3995
Line 3995: static Object* Runtime_SetInGlobalEval(Arguments args) {
Set/ClearPotentialDirectEval?

http://codereview.chromium.org/11563/diff/1/5#newcode4026
Line 4026: return Heap::ToBoolean(eval_origin->IsGlobalObject());
Will this work?  If you put the global object into the scope chain using a with
statement that should be an aliased eval call.

[Mads Ager (chromium), 2008-11-21 12:41:23]

LGTM

http://codereview.chromium.org/11563/diff/25/204
File test/cctest/test-api.cc (right):

http://codereview.chromium.org/11563/diff/25/204#newcode4021
Line 4021: THREADED_TEST(Eval) {
Could you rewrite as many as these tests as possible as mjsunit tests?  They are
easier to read and it seems to me that all of these tests except for one could
be mjsunit tests.

[Mads Ager (chromium), 2008-11-21 14:52:30]

Actually, I think we need to revert this change.  I didn't realize before now,
but this adds an extra runtime call for each call of a javascript function.  We
need to come up with something better.