android_external_v8/src, branch stable/cm-11.0 Unnamed repository; edit this file 'description' to name the repository. Merge upstream V8 change to fix HasRealIndexedProperty. 2013-07-12T16:40:45+00:00 Selim Gurun sgurun@google.com 2013-07-12T16:35:29+00:00 f53dc205fc8ac4f3e92cb554eb15ef0588667a4b Bug: 9720511 Change-Id: I731b49b1bfbeeb9927fb995342f9e7454ce834dc 
Bug: 9720511
Change-Id: I731b49b1bfbeeb9927fb995342f9e7454ce834dc
Use kernel cacheflush for large sizes on MIPS. 2013-05-06T23:10:07+00:00 Paul Lind paul.lind@imgtec.com 2013-05-06T23:09:29+00:00 3ee90b0f201427268e35d86d7ad95e45aa7491a8 This is a workaround to a deserializer bug. The bug was exposed with a recent optimization to use user-mode cache-flushing on MIPS. To reduce risk, we're doing a workaround in mips-specific code so that other arch's cannot be affected. The deserializer does this FlushICache: CPU::FlushICache(last_object_address_, Page::kPageSize); However, that region includes OS guard-pages with no access privilege. The MIPS kernel cacheflush routines work OK in this case, but the Bionic cacheflush recently enabled user-mode flushing using the synci instruction, which causes a segfault on MIPS when the guard pages are reached. (change I48fd6f2b0cbe80c3cd90f453ced97a2f154f7ad3) The workaround just reverts to the kernel flush when the size is Page::kPageSize or bigger. A better fix would be to alter the deserializer so that only the executable pages are flushed: CPU::FlushICache(last_object_address_, isolate_->memory_allocator()->CodePageAreaSize()); However, that changes common code for all supported architectures. There is no evidence that this bug affects the other arch's, so we are doing a MIPS-specific workaround. Change-Id: I30b62eb579feab1453d3ae85a5fb9b408f91756b Signed-off-by: Paul Lind <paul.lind@imgtec.com> 
This is a workaround to a deserializer bug. The bug was exposed
with a recent optimization to use user-mode cache-flushing on
MIPS. To reduce risk, we're doing a workaround in mips-specific
code so that other arch's cannot be affected.

The deserializer does this FlushICache:

  CPU::FlushICache(last_object_address_, Page::kPageSize);

However, that region includes OS guard-pages with no access
privilege. The MIPS kernel cacheflush routines work OK in this
case, but the Bionic cacheflush recently enabled user-mode flushing
using the synci instruction, which causes a segfault on MIPS when
the guard pages are reached.
(change I48fd6f2b0cbe80c3cd90f453ced97a2f154f7ad3)

The workaround just reverts to the kernel flush when the size
is Page::kPageSize or bigger.

A better fix would be to alter the deserializer so that only the
executable pages are flushed:
  CPU::FlushICache(last_object_address_,
                   isolate_->memory_allocator()->CodePageAreaSize());

However, that changes common code for all supported architectures.
There is no evidence that this bug affects the other arch's, so we
are doing a MIPS-specific workaround.

Change-Id: I30b62eb579feab1453d3ae85a5fb9b408f91756b
Signed-off-by: Paul Lind <paul.lind@imgtec.com>
Fix ARM hardfloat detection in linux 2012-11-03T04:54:11+00:00 Andrew Hsieh andrewhsieh@google.com 2012-11-03T04:54:11+00:00 af537b8206a515f942e5ee0338113bd485b55eb7 See http://code.google.com/p/v8/issues/detail?id=2140 https://chromiumcodereview.appspot.com/10713009 The original code fails to detect at run-time when compiled with GCC 4.7, because the undefined behavior of casting void to double, and the r0/r1 clobbered in assembly code isn't reaching the use when compared to 1.0 in VFP reg. In summary, the old code is incorrect and overkill, and the new code fix it. Change-Id: I6b63a4f9789e08089368e431a5553f482400725a 
See
  http://code.google.com/p/v8/issues/detail?id=2140
  https://chromiumcodereview.appspot.com/10713009

The original code fails to detect at run-time when compiled
with GCC 4.7, because the undefined behavior of casting
void to double, and the r0/r1 clobbered in assembly code
isn't reaching the use when compared to 1.0 in VFP reg.
In summary, the old code is incorrect and overkill, and the
new code fix it.

Change-Id: I6b63a4f9789e08089368e431a5553f482400725a
Revert "Android ucontext refinements (upstream backport)". 2012-10-17T17:03:16+00:00 David 'Digit' Turner digit@android.com 2012-10-17T17:01:49+00:00 1bbccaca327b1fed6859a62f6d3ed75a75e4856a This creates build breakages in the internal Android tree. Will investigate later. Original patch: https://android-review.googlesource.com/#/c/38852/ Change-Id: I902021fa56b959f1f035f4d448a43483328debd8 
This creates build breakages in the internal Android tree.
Will investigate later.

Original patch: https://android-review.googlesource.com/#/c/38852/

Change-Id: I902021fa56b959f1f035f4d448a43483328debd8
Android ucontext refinements (upstream backport). 2012-10-17T12:58:16+00:00 David 'Digit' Turner digit@android.com 2012-07-12T16:51:00+00:00 8a0da0b040a743677f20f29548ff3380ffb9a5ed This is a forward-compatible patch to avoid naming/type conflicts in the near future, when the Android C library will be updated to provide mcontext_t/ucontext_t properly. Note that this changes the naming convention used to access the register values from mcontext_t, to follow current ARM GLibc (this is also the convention being adopted by Android). + Android provides gettid() (all API levels) so use it instead of relying on syscall(). This is a backport of http://code.google.com/p/v8/source/detail?r=12250 See http://code.google.com/p/android/issues/detail?id=34784 for more context. Change-Id: Id1fe5489eea8a51d0210858bf91132191f967822 
This is a forward-compatible patch to avoid naming/type conflicts in
the near future, when the Android C library will be updated to provide
mcontext_t/ucontext_t properly.

Note that this changes the naming convention used to access the
register values from mcontext_t, to follow current ARM GLibc (this
is also the convention being adopted by Android).

+ Android provides gettid() (all API levels) so use it instead of
  relying on syscall().

This is a backport of http://code.google.com/p/v8/source/detail?r=12250

See http://code.google.com/p/android/issues/detail?id=34784 for more
context.

Change-Id: Id1fe5489eea8a51d0210858bf91132191f967822
Merge V8 at 3.9.24.30 2012-06-08T13:45:43+00:00 Ben Murdoch benm@google.com 2012-06-08T13:45:43+00:00 7d088e4f8fcf1ffd0817802d219d55b6122a4736 Bug: 6605887 Change-Id: I10a55d1fc257903505a5c8a5933174e1d0fe6052 
Bug: 6605887
Change-Id: I10a55d1fc257903505a5c8a5933174e1d0fe6052
Remove Android specific LUMP_OF_MEMORY. 2012-06-07T17:40:43+00:00 Ben Murdoch benm@google.com 2012-06-07T17:38:12+00:00 a28cdeeef5f062a40c344cbbbcf296873391bc58 The existing 128KB is too small - the default 1MB is more appropriate. This is being tracked in upstream CL http://codereview.chromium.org/10541053/ Bug: 6605887 Change-Id: Iae291aaeab1e7c48b2570ab246769bad391edf56 
The existing 128KB is too small - the default 1MB is more
appropriate.

This is being tracked in upstream CL
http://codereview.chromium.org/10541053/

Bug: 6605887
Change-Id: Iae291aaeab1e7c48b2570ab246769bad391edf56
Merge V8 at 3.9.24.29 2012-05-24T10:17:13+00:00 Ben Murdoch benm@google.com 2012-05-24T10:17:13+00:00 4768e9d22b13edbae2ae177109420028748bf872 Bug: 6552361 Change-Id: Ic75bc91849fed7b20846e9c10e2c1ffcc5239da5 
Bug: 6552361
Change-Id: Ic75bc91849fed7b20846e9c10e2c1ffcc5239da5
Merge V8 at 3.9.24.28 2012-05-21T13:52:42+00:00 Ben Murdoch benm@google.com 2012-05-21T13:52:42+00:00 5710ceac03e2cf7a164ad7393b5a6b6114ea45e6 Bug: 6519007 / 6517949 / 6518772 Change-Id: Idc3e1297aeafa0274fa9fc321e91b774091d1b50 
Bug: 6519007 / 6517949 / 6518772
Change-Id: Idc3e1297aeafa0274fa9fc321e91b774091d1b50
Merge V8 at 3.9.24.25 2012-05-17T10:27:31+00:00 Ben Murdoch benm@google.com 2012-05-17T09:59:30+00:00 84774f4b8cb56f14184c96dd08fb2ae119d986e0 Bug: 6510301 Change-Id: I2e1efb9b3420fd4c72927c9948546e1de16865fc 
Bug: 6510301
Change-Id: I2e1efb9b3420fd4c72927c9948546e1de16865fc