Always force a major GC after each input in v8_serialized_script_value_fuzzer.

Always force a major GC after each input in v8_serialized_script_value_fuzzer.

This causes a significant slowdown, but there have been persistent issues with
trying to be more clever and do infrequent collections (and doing so increases
cross-contamination between inputs).

This will have both V8 and Oilpan do one major GC each time (note that if there
is a chain of persistent objects, some objects may not be freed until a later
cycle).

BUG=659357
Review-Url: https://codereview.chromium.org/2658653003
Cr-Commit-Position: refs/heads/master@{#446320}
Committed: https://chromium.googlesource.com/chromium/src/+/bd795a387b29430c851f367b2b7f3015f47a4ff6

[jbroman, 2017-01-25 20:36:20]

The CQ bit was checked by jbroman@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2017-01-25 20:36:54]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[jbroman, 2017-01-25 20:42:16]

jbroman@chromium.org changed reviewers:
+ mmoroz@chromium.org

[jbroman, 2017-01-25 20:42:16]

PTAL.

This is a big hammer and will slow down v8_serialized_script_value_fuzzer, but
will hopefully (knock on wood) solve bug 659357 (it makes the case reported
there hold constant for me).

I suspect that V8/Oilpan just didn't consider the pressure an issue (because
they don't know about the fuzzer's RSS limit), and weren't getting the same
idle-collection signals they get in a full instance of Chrome.

[haraken, 2017-01-25 20:59:23]

Yeah, let's see how it goes. If the performance becomes a problem, we can think
about teaching the pressure to V8/Oilpan. LGTM.

Nit: Maybe can we use V8GCController::collectAllGarbageForTesting instead?

[jbroman, 2017-01-25 21:03:42]

The CQ bit was checked by jbroman@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2017-01-25 21:03:55]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[jbroman, 2017-01-25 21:04:22]

On 2017/01/25 at 20:59:23, haraken wrote:
> Yeah, let's see how it goes. If the performance becomes a problem, we can
think about teaching the pressure to V8/Oilpan. LGTM.
> 
> Nit: Maybe can we use V8GCController::collectAllGarbageForTesting instead?

This is already a pretty big slowdown, which is why I'm compromising at 1 major
GC (each) instead of 5 (which is what V8GCController does).

[haraken, 2017-01-25 21:05:52]

On 2017/01/25 21:04:22, jbroman wrote:
> On 2017/01/25 at 20:59:23, haraken wrote:
> > Yeah, let's see how it goes. If the performance becomes a problem, we can
> think about teaching the pressure to V8/Oilpan. LGTM.
> > 
> > Nit: Maybe can we use V8GCController::collectAllGarbageForTesting instead?
> 
> This is already a pretty big slowdown, which is why I'm compromising at 1
major
> GC (each) instead of 5 (which is what V8GCController does).

Makes sense.

[commit-bot: I haz the power, 2017-01-25 22:17:14]

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

[commit-bot: I haz the power, 2017-01-25 22:17:15]

Dry run: This issue passed the CQ dry run.

[mmoroz, 2017-01-26 11:28:03]

LGTM!

CC'ed ahaas@ as FYI

[jbroman, 2017-01-26 14:20:04]

The CQ bit was checked by jbroman@chromium.org

[jbroman, 2017-01-26 14:20:05]

The patchset sent to the CQ was uploaded after l-g-t-m from haraken@chromium.org
Link to the patchset: https://codereview.chromium.org/2658653003/#ps20001
(title: "correct number in comment")

[commit-bot: I haz the power, 2017-01-26 14:20:22]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[commit-bot: I haz the power, 2017-01-26 14:23:59]

CQ is committing da patch.

Bot data: {"patchset_id": 20001, "attempt_start_ts": 1485440404971600,
"parent_rev": "781262f98757b3e657a9089de7f98b9752482d09", "commit_rev":
"bd795a387b29430c851f367b2b7f3015f47a4ff6"}

[commit-bot: I haz the power, 2017-01-26 14:24:32]

Description was changed from

==========
Always force a major GC after each input in v8_serialized_script_value_fuzzer.

This causes a significant slowdown, but there have been persistent issues with
trying to be more clever and do infrequent collections (and doing so increases
cross-contamination between inputs).

This will have both V8 and Oilpan do one major GC each time (note that if there
is a chain of persistent objects, some objects may not be freed until a later
cycle).

BUG=659357
==========

to

==========
Always force a major GC after each input in v8_serialized_script_value_fuzzer.

This causes a significant slowdown, but there have been persistent issues with
trying to be more clever and do infrequent collections (and doing so increases
cross-contamination between inputs).

This will have both V8 and Oilpan do one major GC each time (note that if there
is a chain of persistent objects, some objects may not be freed until a later
cycle).

BUG=659357

Review-Url: https://codereview.chromium.org/2658653003
Cr-Commit-Position: refs/heads/master@{#446320}
Committed:
https://chromium.googlesource.com/chromium/src/+/bd795a387b29430c851f367b2b7f...
==========

[commit-bot: I haz the power, 2017-01-26 14:24:33]

Committed patchset #2 (id:20001) as
https://chromium.googlesource.com/chromium/src/+/bd795a387b29430c851f367b2b7f...