[interpreter] Clear pending message object on handler entry.

[interpreter] Clear pending message object on handler entry.

This clears the currently pending message object whenever a try-block or
a finally-block is being entered in interpreted code. The intention is
to avoid memory leaks introduced by the message object. Also the message
object is being restored when a finally-block exits.

R=rmcilroy@chromium.org
TEST=cctest/test-heap/MessageObjectLeak
BUG=v8:4674
LOG=n

Committed: https://crrev.com/e814457675d615aaebb14ae723a42c756492f732
Cr-Commit-Position: refs/heads/master@{#33704}

[rmcilroy, 2016-02-01 14:21:46]

lgtm, thanks.

https://codereview.chromium.org/1651993002/diff/40001/test/cctest/interpreter...
File test/cctest/interpreter/test-bytecode-generator.cc (right):

https://codereview.chromium.org/1651993002/diff/40001/test/cctest/interpreter...
test/cctest/interpreter/test-bytecode-generator.cc:4453: /*           */ R(0),
U8(0),                                    //
Nice hack for git cl format - I'll need to remember that one :).

[Michael Starzinger, 2016-02-01 17:39:52]

Description was changed from

==========
[interpreter] Clear pending message object on handler entry.

This clears the currently pending message object whenever a try-block or
a finally-block is being entered in interpreted code. The intention is
to avoid memory leaks introduced by the message object.

R=rmcilroy@chromium.org
TEST=cctest/test-heap/MessageObjectLeak
BUG=v8:4674
LOG=n
==========

to

==========
[interpreter] Clear pending message object on handler entry.

This clears the currently pending message object whenever a try-block or
a finally-block is being entered in interpreted code. The intention is
to avoid memory leaks introduced by the message object. Also the message
object is being restored when a finally-block exits.

R=rmcilroy@chromium.org
TEST=cctest/test-heap/MessageObjectLeak
BUG=v8:4674
LOG=n
==========

[Michael Starzinger, 2016-02-01 18:35:35]

PTAL. I had to extend this to do the whole "save and restore" dance for finally
blocks in this CL, otherwise existing tests started to fail. So this CL now
contains all the message handling in one go.

[rmcilroy, 2016-02-01 20:57:14]

https://codereview.chromium.org/1651993002/diff/80001/src/interpreter/bytecod...
File src/interpreter/bytecode-generator.cc (right):

https://codereview.chromium.org/1651993002/diff/80001/src/interpreter/bytecod...
src/interpreter/bytecode-generator.cc:1193:
builder()->CallRuntime(Runtime::kInterpreterClearPendingMessage, no_reg, 0);
Could we make InterpreterGetPendingMessage also clear the pending message to
avoid having to make two runtime calls in quick succession?

https://codereview.chromium.org/1651993002/diff/80001/test/cctest/cctest.status
File test/cctest/cctest.status (right):

https://codereview.chromium.org/1651993002/diff/80001/test/cctest/cctest.stat...
test/cctest/cctest.status:513: 'test-run-deopt/DeoptExceptionHandlerCatch':
[PASS, FAIL],
Is this now flaky or does it fail only on some configurations?

[Michael Starzinger, 2016-02-02 09:41:43]

Addressed comments.

https://codereview.chromium.org/1651993002/diff/80001/src/interpreter/bytecod...
File src/interpreter/bytecode-generator.cc (right):

https://codereview.chromium.org/1651993002/diff/80001/src/interpreter/bytecod...
src/interpreter/bytecode-generator.cc:1193:
builder()->CallRuntime(Runtime::kInterpreterClearPendingMessage, no_reg, 0);
On 2016/02/01 20:57:14, rmcilroy wrote:
> Could we make InterpreterGetPendingMessage also clear the pending message to
> avoid having to make two runtime calls in quick succession?

Done. I made kInterpreterClearPendingMessage return the original message.

https://codereview.chromium.org/1651993002/diff/80001/test/cctest/cctest.status
File test/cctest/cctest.status (right):

https://codereview.chromium.org/1651993002/diff/80001/test/cctest/cctest.stat...
test/cctest/cctest.status:513: 'test-run-deopt/DeoptExceptionHandlerCatch':
[PASS, FAIL],
On 2016/02/01 20:57:14, rmcilroy wrote:
> Is this now flaky or does it fail only on some configurations?

Not flaky but firing a DCHECK, i.e. PASS in release and FAIL in debug mode.

[rmcilroy, 2016-02-03 11:37:21]

lgtm, thanks.

[Michael Starzinger, 2016-02-03 13:43:21]

The CQ bit was checked by mstarzinger@chromium.org

[Michael Starzinger, 2016-02-03 13:43:21]

The patchset sent to the CQ was uploaded after l-g-t-m from
rmcilroy@chromium.org
Link to the patchset: https://codereview.chromium.org/1651993002/#ps120001
(title: "Rebased.")

[commit-bot: I haz the power, 2016-02-03 13:43:36]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1651993002/120001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1651993002/120001

[commit-bot: I haz the power, 2016-02-03 13:52:37]

Description was changed from

==========
[interpreter] Clear pending message object on handler entry.

This clears the currently pending message object whenever a try-block or
a finally-block is being entered in interpreted code. The intention is
to avoid memory leaks introduced by the message object. Also the message
object is being restored when a finally-block exits.

R=rmcilroy@chromium.org
TEST=cctest/test-heap/MessageObjectLeak
BUG=v8:4674
LOG=n
==========

to

==========
[interpreter] Clear pending message object on handler entry.

This clears the currently pending message object whenever a try-block or
a finally-block is being entered in interpreted code. The intention is
to avoid memory leaks introduced by the message object. Also the message
object is being restored when a finally-block exits.

R=rmcilroy@chromium.org
TEST=cctest/test-heap/MessageObjectLeak
BUG=v8:4674
LOG=n
==========

[commit-bot: I haz the power, 2016-02-03 13:52:38]

Committed patchset #7 (id:120001)

[commit-bot: I haz the power, 2016-02-03 13:52:52]

Description was changed from

==========
[interpreter] Clear pending message object on handler entry.

This clears the currently pending message object whenever a try-block or
a finally-block is being entered in interpreted code. The intention is
to avoid memory leaks introduced by the message object. Also the message
object is being restored when a finally-block exits.

R=rmcilroy@chromium.org
TEST=cctest/test-heap/MessageObjectLeak
BUG=v8:4674
LOG=n
==========

to

==========
[interpreter] Clear pending message object on handler entry.

This clears the currently pending message object whenever a try-block or
a finally-block is being entered in interpreted code. The intention is
to avoid memory leaks introduced by the message object. Also the message
object is being restored when a finally-block exits.

R=rmcilroy@chromium.org
TEST=cctest/test-heap/MessageObjectLeak
BUG=v8:4674
LOG=n

Committed: https://crrev.com/e814457675d615aaebb14ae723a42c756492f732
Cr-Commit-Position: refs/heads/master@{#33704}
==========

[commit-bot: I haz the power, 2016-02-03 13:52:53]

Patchset 7 (id:??) landed as
https://crrev.com/e814457675d615aaebb14ae723a42c756492f732
Cr-Commit-Position: refs/heads/master@{#33704}