[es6] Sloppy functions in block

[es6] Sloppy functions in block

http://www.ecma-international.org/ecma-262/6.0/#sec-block-level-function-declarations-web-legacy-compatibility-semantics

With --harmony-sloppy we now create the 2 bindings as needed for sloppy
functions in blocks.

BUG=v8:4285
LOG=N
R=rossberg@chromium.org, littledan@chromium.org

[arv (Not doing code reviews), 2015-07-09 21:05:05]

I need to add another bailout id because we cannot reuse the id.

[arv (Not doing code reviews), 2015-07-09 21:23:16]

We need to use 2 different unresoved proxies

[rossberg, 2015-07-10 12:18:27]

LGTM modulo comment

https://codereview.chromium.org/1234433002/diff/20001/test/mjsunit/harmony/bl...
File test/mjsunit/harmony/block-function-sloppy.js (right):

https://codereview.chromium.org/1234433002/diff/20001/test/mjsunit/harmony/bl...
test/mjsunit/harmony/block-function-sloppy.js:30: (function TestNoInitIf() {
You have all these tests for the NoInit case, but none for a successful
control-dependent case. Can you add some, too?

[arv (Not doing code reviews), 2015-07-10 12:56:12]

Thanks. I'll add more tests and investigate the error reporting.

https://codereview.chromium.org/1234433002/diff/20001/src/parser.cc
File src/parser.cc (right):

https://codereview.chromium.org/1234433002/diff/20001/src/parser.cc#newcode2230
src/parser.cc:2230: Declare(declaration, DeclarationDescriptor::NORMAL, true,
&var_ok);
I was wondering if this actually reports an error. I'll need to look into it and
add test etc.

https://codereview.chromium.org/1234433002/diff/20001/test/mjsunit/harmony/bl...
File test/mjsunit/harmony/block-function-sloppy.js (right):

https://codereview.chromium.org/1234433002/diff/20001/test/mjsunit/harmony/bl...
test/mjsunit/harmony/block-function-sloppy.js:30: (function TestNoInitIf() {
On 2015/07/10 12:18:27, rossberg wrote:
> You have all these tests for the NoInit case, but none for a successful
> control-dependent case. Can you add some, too?

Doh! I will add.

[arv (Not doing code reviews), 2015-07-10 16:36:10]

Add more tests and discover we do report wrong error

[arv (Not doing code reviews), 2015-07-10 16:37:23]

https://codereview.chromium.org/1234433002/diff/40001/test/mjsunit/harmony/bl...
File test/mjsunit/harmony/block-function-sloppy.js (right):

https://codereview.chromium.org/1234433002/diff/40001/test/mjsunit/harmony/bl...
test/mjsunit/harmony/block-function-sloppy.js:225: assertEquals("Identifier 'y'
has already been declared", ex.message);
We are reporting 'x' here as I suspected. Will have to look into Declare to make
sure we do not report these errors.

[arv (Not doing code reviews), 2015-07-10 16:54:32]

Don't report errors

[arv (Not doing code reviews), 2015-07-10 16:56:33]

PTAL

https://codereview.chromium.org/1234433002/diff/60001/src/parser.cc
File src/parser.cc (right):

https://codereview.chromium.org/1234433002/diff/60001/src/parser.cc#newcode1996
src/parser.cc:1996: bool report_error, bool* ok) {
Turned out that all callers always passed resolve as true so changing this to
report_error required no changes for the callers =D

[arv (Not doing code reviews), 2015-07-10 17:49:07]

That did not quite work...

[arv (Not doing code reviews), 2015-07-10 18:20:00]

Turns out resolve was needed after all

[arv (Not doing code reviews), 2015-07-10 18:42:55]

Now I managed to repro the issue...

[arv (Not doing code reviews), 2015-07-10 18:53:58]

Dan, can you take over this when you get a chance?

[arv (Not doing code reviews), 2015-07-12 19:01:41]

This CL is not going to work. We can only determine if it is OK to add the var
binding as we are done with the function/program, at the same place where we
call CheckConflictingVarDeclarations.

I think we can add the declaration as well leave the function/program but the
assignment is a bit trickier. The assignment should be done to both bindings
when the function is created (and not what I did in this CL). I think that we
have to update the codegen to support this.

Anyway, closing this CL since it is incorrect.