Add regression test for r10451.

Add regression test for r10451.

R=vegorov@chromium.org
TEST=cctest/test-heap/PrototypeTransitionClearing

Committed: https://code.google.com/p/v8/source/detail?r=10455

[Vyacheslav Egorov (Chromium), 2012-01-20 12:17:39]

lgtm

https://chromiumcodereview.appspot.com/9138016/diff/1/test/cctest/test-heap.cc
File test/cctest/test-heap.cc (right):

https://chromiumcodereview.appspot.com/9138016/diff/1/test/cctest/test-heap.c...
test/cctest/test-heap.cc:1613: FACTORY->NewJSArray(1, TENURED);
This loop looks a bit spooky because the fact that NewJSArray returns a handle
is kinda hidden so one would expect that these arrays will die and this will
cause infinite loop (but they don't because they actually are referenced from
the local handles).

Please add a comment about this. Also it might sense to use bigger arrays to
make this test complete faster.

https://chromiumcodereview.appspot.com/9138016/diff/1/test/cctest/test-heap.c...
test/cctest/test-heap.cc:1621:
CHECK(!space->LastPage()->Contains(map->prototype_transitions()->address()));
this check look suspicious. by accident there might be a hole on the first page
large enough for this object. You might consider changing the first loop into 


Handle<JSObject> prototype;

do {
prototype = FACTORY->NewJSArray(N, TENURED);
} while (prototype is not on the second page);

[Michael Starzinger, 2012-01-20 12:41:53]

Added new patch set.

https://chromiumcodereview.appspot.com/9138016/diff/1/test/cctest/test-heap.cc
File test/cctest/test-heap.cc (right):

https://chromiumcodereview.appspot.com/9138016/diff/1/test/cctest/test-heap.c...
test/cctest/test-heap.cc:1613: FACTORY->NewJSArray(1, TENURED);
On 2012/01/20 12:17:39, Vyacheslav Egorov wrote:
> This loop looks a bit spooky because the fact that NewJSArray returns a handle
> is kinda hidden so one would expect that these arrays will die and this will
> cause infinite loop (but they don't because they actually are referenced from
> the local handles).
> 
> Please add a comment about this. Also it might sense to use bigger arrays to
> make this test complete faster.

Done. Now unnecessary due to second change.

https://chromiumcodereview.appspot.com/9138016/diff/1/test/cctest/test-heap.c...
test/cctest/test-heap.cc:1621:
CHECK(!space->LastPage()->Contains(map->prototype_transitions()->address()));
On 2012/01/20 12:17:39, Vyacheslav Egorov wrote:
> this check look suspicious. by accident there might be a hole on the first
page
> large enough for this object. You might consider changing the first loop into 
> 
> 
> Handle<JSObject> prototype;
> 
> do {
> prototype = FACTORY->NewJSArray(N, TENURED);
> } while (prototype is not on the second page);

Done. Yes, that is much simpler and I can arbitrarily increase the size of the
prototype without worrying about holes in the free-list. That makes the test
complete much faster. Nice idea!