Make all global private symbols own symbols.

Make all global private symbols own symbols.

R=arv@chromium.org, rossberg@chromium.org, wingo@igalia.com, yangguo@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=23252

[Dmitry Lomov (no reviews), 2014-08-18 15:12:48]

PTAL, this is a step towards making all private symbols own symbols. 
There is a lot to bikeshed in this CL :)

[rossberg, 2014-08-19 12:04:20]

Hm, can't we do the transition in one CL? It seems all you need to add is
getting rid of the separate flag again, and fix the one place where we currently
lookup a private property across prototypes (which is trivial, according to
Yang). That would also avoid the redundant renames of GLOBAL_PRIVATE to
GLOBAL_PRIVATE_OWN in this CL.

[rossberg, 2014-08-19 12:04:31]

https://codereview.chromium.org/486763002/diff/1/src/macros.py
File src/macros.py (right):

https://codereview.chromium.org/486763002/diff/1/src/macros.py#newcode177
src/macros.py:177: macro IS_PRIVATE_OWN_UNDEFINED(obj, sym) =
(IS_UNDEFINED(obj[sym]));
How about HAS_DEFINED_PRIVATE, for more consistency?

[Dmitry Lomov (no reviews), 2014-08-19 13:17:52]

On 2014/08/19 12:04:20, rossberg wrote:
> Hm, can't we do the transition in one CL? It seems all you need to add is
> getting rid of the separate flag again, and fix the one place where we
currently
> lookup a private property across prototypes (which is trivial, according to
> Yang). That would also avoid the redundant renames of GLOBAL_PRIVATE to
> GLOBAL_PRIVATE_OWN in this CL.

If you have time to the whole thing in one go, feel free :)

[Dmitry Lomov (no reviews), 2014-08-19 13:18:01]

On 2014/08/19 12:04:20, rossberg wrote:
> Hm, can't we do the transition in one CL? It seems all you need to add is
> getting rid of the separate flag again, and fix the one place where we
currently
> lookup a private property across prototypes (which is trivial, according to
> Yang). That would also avoid the redundant renames of GLOBAL_PRIVATE to
> GLOBAL_PRIVATE_OWN in this CL.

If you have time to the whole thing in one go, feel free :)

[Dmitry Lomov (no reviews), 2014-08-19 13:29:56]

Renamed per CR feedback

https://codereview.chromium.org/486763002/diff/1/src/macros.py
File src/macros.py (right):

https://codereview.chromium.org/486763002/diff/1/src/macros.py#newcode177
src/macros.py:177: macro IS_PRIVATE_OWN_UNDEFINED(obj, sym) =
(IS_UNDEFINED(obj[sym]));
On 2014/08/19 12:04:31, rossberg wrote:
> How about HAS_DEFINED_PRIVATE, for more consistency?

Done.

[wingo, 2014-08-20 15:32:53]

https://codereview.chromium.org/486763002/diff/20001/src/macros.py
File src/macros.py (right):

https://codereview.chromium.org/486763002/diff/20001/src/macros.py#newcode173
src/macros.py:173: macro GLOBAL_PRIVATE_OWN(name) =
(%CreateGlobalPrivateOwnSymbol(name));
Q: global private symbols are cross-realm, and normal private symbols are not?

https://codereview.chromium.org/486763002/diff/20001/src/string-iterator.js
File src/string-iterator.js (right):

https://codereview.chromium.org/486763002/diff/20001/src/string-iterator.js#n...
src/string-iterator.js:49: if (s === stringEndOfIterationMarker) {
Wouldn't stringEndOfIterationMarker be per-realm?  I always get this stuff
wrong.

[arv (Not doing code reviews), 2014-08-20 15:39:27]

https://codereview.chromium.org/486763002/diff/20001/src/string-iterator.js
File src/string-iterator.js (right):

https://codereview.chromium.org/486763002/diff/20001/src/string-iterator.js#n...
src/string-iterator.js:49: if (s === stringEndOfIterationMarker) {
On 2014/08/20 15:32:53, wingo wrote:
> Wouldn't stringEndOfIterationMarker be per-realm?  I always get this stuff
> wrong.

You are right. This is not going to work.

I don't see why we cannot continue to use undefined here?

[Dmitry Lomov (no reviews), 2014-08-20 17:01:37]

https://codereview.chromium.org/486763002/diff/20001/src/string-iterator.js
File src/string-iterator.js (right):

https://codereview.chromium.org/486763002/diff/20001/src/string-iterator.js#n...
src/string-iterator.js:49: if (s === stringEndOfIterationMarker) {
On 2014/08/20 15:39:26, arv wrote:
> On 2014/08/20 15:32:53, wingo wrote:
> > Wouldn't stringEndOfIterationMarker be per-realm?  I always get this stuff
> > wrong.
> 
> You are right. This is not going to work.

Thanks, good catch, I haven't thought of that.
> 
> I don't see why we cannot continue to use undefined here?

It's because I have changed the HAS_PRIVATE above to essentially
  iterator[stringIteratorIteratedStringSymbol] === undefined
There are two things I could do:
1) Either keep using %HasOwnProperty above (runtime call)
2) Or use a different symbol for branding (string iterators bigger by 1 field)

Opinions?

[arv (Not doing code reviews), 2014-08-20 18:07:07]

https://codereview.chromium.org/486763002/diff/20001/src/string-iterator.js
File src/string-iterator.js (right):

https://codereview.chromium.org/486763002/diff/20001/src/string-iterator.js#n...
src/string-iterator.js:49: if (s === stringEndOfIterationMarker) {
On 2014/08/20 17:01:37, Dmitry Lomov (chromium) wrote:
> On 2014/08/20 15:39:26, arv wrote:
> > On 2014/08/20 15:32:53, wingo wrote:
> > > Wouldn't stringEndOfIterationMarker be per-realm?  I always get this stuff
> > > wrong.
> > 
> > You are right. This is not going to work.
> 
> Thanks, good catch, I haven't thought of that.
> > 
> > I don't see why we cannot continue to use undefined here?
> 
> It's because I have changed the HAS_PRIVATE above to essentially
>   iterator[stringIteratorIteratedStringSymbol] === undefined
> There are two things I could do:
> 1) Either keep using %HasOwnProperty above (runtime call)
> 2) Or use a different symbol for branding (string iterators bigger by 1 field)
> 
> Opinions? 
> 

In this case and for the ArrayIterator we can check
stringIteratorNextIndexSymbol instead. We set this to 0 in CreateStringIterator
so the "brand" check can be changed to !IS_UNDEFINED(GET_PRIVATE(iterator,
stringIteratorNextIndexSymbol)), which I goes is what HAS_DEFINED_PRIVATE does.

[Dmitry Lomov (no reviews), 2014-08-20 18:17:50]

On 2014/08/20 18:07:07, arv wrote:
> https://codereview.chromium.org/486763002/diff/20001/src/string-iterator.js
> File src/string-iterator.js (right):
> 
>
https://codereview.chromium.org/486763002/diff/20001/src/string-iterator.js#n...
> src/string-iterator.js:49: if (s === stringEndOfIterationMarker) {
> On 2014/08/20 17:01:37, Dmitry Lomov (chromium) wrote:
> > On 2014/08/20 15:39:26, arv wrote:
> > > On 2014/08/20 15:32:53, wingo wrote:
> > > > Wouldn't stringEndOfIterationMarker be per-realm?  I always get this
stuff
> > > > wrong.
> > > 
> > > You are right. This is not going to work.
> > 
> > Thanks, good catch, I haven't thought of that.
> > > 
> > > I don't see why we cannot continue to use undefined here?
> > 
> > It's because I have changed the HAS_PRIVATE above to essentially
> >   iterator[stringIteratorIteratedStringSymbol] === undefined
> > There are two things I could do:
> > 1) Either keep using %HasOwnProperty above (runtime call)
> > 2) Or use a different symbol for branding (string iterators bigger by 1
field)
> > 
> > Opinions? 
> > 
> 
> In this case and for the ArrayIterator we can check
> stringIteratorNextIndexSymbol instead. We set this to 0 in
CreateStringIterator
> so the "brand" check can be changed to !IS_UNDEFINED(GET_PRIVATE(iterator,
> stringIteratorNextIndexSymbol)), which I goes is what HAS_DEFINED_PRIVATE
does.

Right, thanks, will do.

[Dmitry Lomov (no reviews), 2014-08-20 18:45:41]

Please take another look

[arv (Not doing code reviews), 2014-08-20 18:51:28]

LGTM

[Dmitry Lomov (no reviews), 2014-08-20 18:55:16]

On 2014/08/20 18:51:28, arv wrote:
> LGTM

wait scratch that this was a wrong patch, sorry, because I haven't renamed
things like Andreas asked. 
I guess I still suffer from that cold :)

[Dmitry Lomov (no reviews), 2014-08-20 19:05:12]

ok, should be all good now.

[arv (Not doing code reviews), 2014-08-20 19:12:29]

LGTM

That is better

[Dmitry Lomov (no reviews), 2014-08-20 19:22:08]

Committed patchset #5 manually as 23252 (presubmit successful).