Ensure reducing the length of an array doesn't make it go holey.

Ensure reducing the length of an array doesn't make it go holey.

Also only transition and/or change anything to the backing store if we are
actually going to delete anything.


BUG=

Committed: https://code.google.com/p/v8/source/detail?r=12840

[Toon Verwaest, 2012-10-31 16:40:15]

PTAL

[danno, 2012-10-31 18:23:11]

LGTM if you beef up the tests as indicated

https://chromiumcodereview.appspot.com/11358011/diff/1/test/mjsunit/elements-...
File test/mjsunit/elements-length-no-holey.js (right):

https://chromiumcodereview.appspot.com/11358011/diff/1/test/mjsunit/elements-...
test/mjsunit/elements-length-no-holey.js:32:
assertFalse(%HasFastHoleyElements(a));
assertTrue(%FastSmiElements(a));

https://chromiumcodereview.appspot.com/11358011/diff/1/test/mjsunit/regress/r...
File test/mjsunit/regress/regress-delete-empty-double.js (right):

https://chromiumcodereview.appspot.com/11358011/diff/1/test/mjsunit/regress/r...
test/mjsunit/regress/regress-delete-empty-double.js:30: delete a[0];
 assertTrue(%HasFastDoubleElements(a));
 assertTrue(%HasFastHoleyElements(a));

[Toon Verwaest, 2012-10-31 18:27:28]

https://chromiumcodereview.appspot.com/11358011/diff/1/test/mjsunit/regress/r...
File test/mjsunit/regress/regress-delete-empty-double.js (right):

https://chromiumcodereview.appspot.com/11358011/diff/1/test/mjsunit/regress/r...
test/mjsunit/regress/regress-delete-empty-double.js:30: delete a[0];
Are you sure this should be holey? In my current implementation they stay
non-holey given that nothing was actually deleted.

On 2012/10/31 18:23:11, danno wrote:
>  assertTrue(%HasFastDoubleElements(a));
>  assertTrue(%HasFastHoleyElements(a));

[danno, 2012-10-31 20:15:09]

https://chromiumcodereview.appspot.com/11358011/diff/1/test/mjsunit/regress/r...
File test/mjsunit/regress/regress-delete-empty-double.js (right):

https://chromiumcodereview.appspot.com/11358011/diff/1/test/mjsunit/regress/r...
test/mjsunit/regress/regress-delete-empty-double.js:30: delete a[0];
You're right. How about a second case that verifies that deleting element 0 of a
1-element array does make it holey?

On 2012/10/31 18:27:28, Toon Verwaest wrote:
> Are you sure this should be holey? In my current implementation they stay
> non-holey given that nothing was actually deleted.
> 
> On 2012/10/31 18:23:11, danno wrote:
> >  assertTrue(%HasFastDoubleElements(a));
> >  assertTrue(%HasFastHoleyElements(a));
>

[Toon Verwaest, 2012-11-02 09:47:25]

Addressed comments.