Revert "Revert "Fix a bug in scope analysis.""

Revert "Revert "Fix a bug in scope analysis.""

Reapply r8783 with an additional fix.

Because the preparser and parser do not use the same scope analysis to
determine if a function can be lazily compiled, the parser can have false
positives.  Rather than treating this as a parse error, treat the preparser
as authoritative and eagerly compile the function.

R=lrn@chromium.org
BUG=
TEST=

Committed: http://code.google.com/p/v8/source/detail?r=8797

[Lasse Reichstein, 2011-08-03 09:05:17]

LGTM

http://codereview.chromium.org/7565003/diff/3001/src/parser.cc
File src/parser.cc (right):

http://codereview.chromium.org/7565003/diff/3001/src/parser.cc#newcode3737
src/parser.cc:3737: is_lazily_compiled = false;
This handles false positives (from the parser's perspective) only. It also
affects false negatives, if they can occur.  It will completely block all
further laziness, since the ScriptDataImpl will never match (and thereby skip)
the unused function entry.

That's probably fine for now.

[Kevin Millikin (Chromium), 2011-08-03 09:09:49]

http://codereview.chromium.org/7565003/diff/3001/src/parser.cc
File src/parser.cc (right):

http://codereview.chromium.org/7565003/diff/3001/src/parser.cc#newcode3737
src/parser.cc:3737: is_lazily_compiled = false;
On 2011/08/03 09:05:17, Lasse Reichstein wrote:
> This handles false positives (from the parser's perspective) only. It also
> affects false negatives, if they can occur.  It will completely block all
> further laziness, since the ScriptDataImpl will never match (and thereby skip)
> the unused function entry.
> 
> That's probably fine for now.

Thanks for pointing that out, I hadn't realized it.  I want to commit this as
is, but we should clean it up soon.