Fix OOM bug 3976.

Fix OOM bug 3976.

Also introduce --trace-fragmentation-verbose, and fix --always-compact.

R=ulan@chromium.org
BUG=v8:3976
LOG=y

Committed: https://crrev.com/4c806802b5e839e29e6febbd68163adda37f1ad7
Cr-Commit-Position: refs/heads/master@{#27414}

[ulan, 2015-03-20 10:56:54]

https://codereview.chromium.org/1024823002/diff/20001/src/heap/mark-compact.cc
File src/heap/mark-compact.cc (right):

https://codereview.chromium.org/1024823002/diff/20001/src/heap/mark-compact.c...
src/heap/mark-compact.cc:648: static const int kMaxMaxEvacuationCandidates =
10000;
Does this help? I think 1000 pages (1GB) should be enough.

https://codereview.chromium.org/1024823002/diff/20001/src/heap/mark-compact.c...
src/heap/mark-compact.cc:3062: if (space->HasEmergencyMemory() ||
space->CanExpand()) {
Good catch!

https://codereview.chromium.org/1024823002/diff/20001/src/hydrogen.cc
File src/hydrogen.cc (right):

https://codereview.chromium.org/1024823002/diff/20001/src/hydrogen.cc#newcode...
src/hydrogen.cc:3359: function_state_ = &initial_function_state_;
Nit: this probably belongs to another CL, but I am fine with landing it here.

https://codereview.chromium.org/1024823002/diff/20001/test/mjsunit/regress/re...
File test/mjsunit/regress/regress-3976.js (right):

https://codereview.chromium.org/1024823002/diff/20001/test/mjsunit/regress/re...
test/mjsunit/regress/regress-3976.js:28: table = [];
Can we get faster test by setting max_old_space_size to lower value and
decreasing 1000000 below?

[Erik Corry, 2015-03-20 11:39:22]

erik.corry@gmail.com changed reviewers:
+ erik.corry@gmail.com

[Erik Corry, 2015-03-20 11:39:23]

https://codereview.chromium.org/1024823002/diff/20001/src/heap/mark-compact.cc
File src/heap/mark-compact.cc (right):

https://codereview.chromium.org/1024823002/diff/20001/src/heap/mark-compact.c...
src/heap/mark-compact.cc:648: static const int kMaxMaxEvacuationCandidates =
10000;
On 2015/03/20 10:56:53, ulan wrote:
> Does this help? I think 1000 pages (1GB) should be enough.

That rather depends on the total heap size.  Not everyone runs with the limits
that Chromium has.  It won't normally make a difference without --always-compact
though.

https://codereview.chromium.org/1024823002/diff/20001/src/hydrogen.cc
File src/hydrogen.cc (right):

https://codereview.chromium.org/1024823002/diff/20001/src/hydrogen.cc#newcode...
src/hydrogen.cc:3359: function_state_ = &initial_function_state_;
On 2015/03/20 10:56:54, ulan wrote:
> Nit: this probably belongs to another CL, but I am fine with landing it here.

git cl won't let me omit this change for some reason

https://codereview.chromium.org/1024823002/diff/20001/test/mjsunit/regress/re...
File test/mjsunit/regress/regress-3976.js (right):

https://codereview.chromium.org/1024823002/diff/20001/test/mjsunit/regress/re...
test/mjsunit/regress/regress-3976.js:28: table = [];
On 2015/03/20 10:56:54, ulan wrote:
> Can we get faster test by setting max_old_space_size to lower value and
> decreasing 1000000 below?

Good idea.  Reducing by a factor of 20 with a max old space size of 50.  That
seems to work on both 32 and 64 bit, and it fails without the &&->|| change.

[ulan, 2015-03-20 11:43:24]

lgtm!

[Erik Corry Chromium.org, 2015-03-20 11:48:07]

The CQ bit was checked by erikcorry@chromium.org

[Erik Corry Chromium.org, 2015-03-20 11:48:08]

The patchset sent to the CQ was uploaded after l-g-t-m from ulan@chromium.org
Link to the patchset: https://codereview.chromium.org/1024823002/#ps60001
(title: "Revert kMaxMax change, replace with warning")

[commit-bot: I haz the power, 2015-03-20 11:48:25]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1024823002/60001

[commit-bot: I haz the power, 2015-03-20 11:48:26]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-03-20 11:48:28]

CLs for remote refs other than refs/pending/heads/master must contain NOTRY=true
in order for the CQ to process them

[Erik Corry Chromium.org, 2015-03-20 12:21:51]

The CQ bit was checked by erikcorry@chromium.org

[Erik Corry Chromium.org, 2015-03-20 12:21:51]

The patchset sent to the CQ was uploaded after l-g-t-m from ulan@chromium.org
Link to the patchset: https://codereview.chromium.org/1024823002/#ps100001
(title: "Up heap limit on test")

[commit-bot: I haz the power, 2015-03-20 12:22:01]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1024823002/100001

[commit-bot: I haz the power, 2015-03-20 12:50:13]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-03-20 12:50:15]

Try jobs failed on following builders:
  v8_linux64_rel on tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_linux64_rel/builds/4148)

[Jakob Kummerow, 2015-03-21 13:38:54]

jkummerow@chromium.org changed reviewers:
+ jkummerow@chromium.org

[Jakob Kummerow, 2015-03-21 13:38:55]

DBC.

Since this CL clearly targets a bug, it should have a proper BUG=v8:3976 line,
so that (1) there's a convenient link to click, and (2) bugdroid autogenerates
appropriate comments.

https://codereview.chromium.org/1024823002/diff/100001/test/mjsunit/mjsunit.s...
File test/mjsunit/mjsunit.status (right):

https://codereview.chromium.org/1024823002/diff/100001/test/mjsunit/mjsunit.s...
test/mjsunit/mjsunit.status:141: 'regress/regress-3976': [PASS, NO_VARIANTS,
['arch == arm or arch == android_arm or arch == android_arm64 or arch == mipsel
or arch == mips64el or arch == mips', TIMEOUT]],
Consider s/TIMEOUT/SKIP/ to save time on the bots. Or does it produce useful
coverage to have the test run but timeout?

[Erik Corry Chromium.org, 2015-03-24 12:44:22]

Had to modify an assert to allow emergency allocations, PTAL.

https://codereview.chromium.org/1024823002/diff/100001/test/mjsunit/mjsunit.s...
File test/mjsunit/mjsunit.status (right):

https://codereview.chromium.org/1024823002/diff/100001/test/mjsunit/mjsunit.s...
test/mjsunit/mjsunit.status:141: 'regress/regress-3976': [PASS, NO_VARIANTS,
['arch == arm or arch == android_arm or arch == android_arm64 or arch == mipsel
or arch == mips64el or arch == mips', TIMEOUT]],
On 2015/03/21 13:38:55, Jakob wrote:
> Consider s/TIMEOUT/SKIP/ to save time on the bots. Or does it produce useful
> coverage to have the test run but timeout?

Done

[ulan, 2015-03-24 12:49:16]

> Had to modify an assert to allow emergency allocations, PTAL.
LGTM

[Erik Corry Chromium.org, 2015-03-24 13:33:54]

The CQ bit was checked by erikcorry@chromium.org

[commit-bot: I haz the power, 2015-03-24 13:34:08]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1024823002/140001

[commit-bot: I haz the power, 2015-03-24 13:53:45]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-03-24 13:53:46]

Try jobs failed on following builders:
  v8_win64_rel on tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_win64_rel/builds/3215)

[Erik Corry Chromium.org, 2015-03-24 14:29:10]

The CQ bit was checked by erikcorry@chromium.org

[Erik Corry Chromium.org, 2015-03-24 14:29:10]

The patchset sent to the CQ was uploaded after l-g-t-m from ulan@chromium.org
Link to the patchset: https://codereview.chromium.org/1024823002/#ps160001
(title: "Fix signed/unsigned comparison")

[commit-bot: I haz the power, 2015-03-24 14:29:22]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1024823002/160001

[commit-bot: I haz the power, 2015-03-24 15:02:27]

Committed patchset #9 (id:160001)

[commit-bot: I haz the power, 2015-03-24 15:02:34]

Patchset 9 (id:??) landed as
https://crrev.com/4c806802b5e839e29e6febbd68163adda37f1ad7
Cr-Commit-Position: refs/heads/master@{#27414}