[inspector] better stacks for promises

[inspector] better stacks for promises

- we should always set creation async stack if it's available regardless existing of current parent async stack,
- we should cleanup parent link iff there is no creation and schedule async stack for parent.

Let's consider example: Promise.resolve().then(x => x).then(x => x), there is three promises which will call following instrumentation:
1) created #1 (Promise.resolve()) - collected stack #1
2) scheduled #1 - collected stack #2
3) created #2 with #1 as parent (first .then) - collected stack #3
4) created #3 with #2 as parent (first .then) - collected stack #4
5) started #2 - use stack #2 as scheduled
6) scheduled #2 - collected stack #6
7) finished #2
8) started #3 - use stack #6 as scheduled 
9) scheduled #3 - collected stack #7
10) finished #3

If we collect stacks between step 4 and 5, it's possible to collect scheduled stack #2 but still have creation stack for #2 - stack #3 - so we always need to add creation event if scheduled is collected.

If we collect stacks between created and scheduled we should not remove parent link even if parent was not scheduled yet.

BUG=v8:6189
R=dgozman@chromium.org

Review-Url: https://codereview.chromium.org/2844753002
Cr-Commit-Position: refs/heads/master@{#44990}
Committed: https://chromium.googlesource.com/v8/v8/+/f2bd913cd459854894447f5422a953d23fa4e7ba

[kozy, 2017-04-26 17:41:39]

Dmitry, please take a look.

https://codereview.chromium.org/2844753002/diff/1/test/inspector/debugger/col...
File test/inspector/debugger/collect-old-async-call-chains-expected.txt (left):

https://codereview.chromium.org/2844753002/diff/1/test/inspector/debugger/col...
test/inspector/debugger/collect-old-async-call-chains-expected.txt:174: actual
async chain len: 2
it was an empty async parent:
parent: {
callFrames: [],
description: 'Promise.resolve'
}

[kozy, 2017-04-26 17:41:54]

Description was changed from

==========
[inspector] better stacks for promises

- we should always setup creation async stack if it's available regardless
existing of current parent async stack,
- we should cleanup parent link iff there is no creation and schedule async
stack for parent.

BUG=v8:6189
R=dgozman@chromium.org
==========

to

==========
[inspector] better stacks for promises

- we should always set creation async stack if it's available regardless
existing of current parent async stack,
- we should cleanup parent link iff there is no creation and schedule async
stack for parent.

BUG=v8:6189
R=dgozman@chromium.org
==========

[dgozman, 2017-04-27 17:26:35]

I failed to understand the case where previous code worked incorrectly. Can we
have a better description? Perhaps with a snippet?

https://codereview.chromium.org/2844753002/diff/1/src/inspector/v8-debugger.cc
File src/inspector/v8-debugger.cc (right):

https://codereview.chromium.org/2844753002/diff/1/src/inspector/v8-debugger.c...
src/inspector/v8-debugger.cc:1053: // scheduled task#2 - stack#5
scheduled after started?

https://codereview.chromium.org/2844753002/diff/1/test/inspector/debugger/pro...
File test/inspector/debugger/promise-chain-when-limit-hit-expected.txt (right):

https://codereview.chromium.org/2844753002/diff/1/test/inspector/debugger/pro...
test/inspector/debugger/promise-chain-when-limit-hit-expected.txt:119:
inspector.setMaxAsyncTaskStacks(6)
Why this one is shorter than previous?

[kozy, 2017-04-27 18:35:11]

Description was changed from

==========
[inspector] better stacks for promises

- we should always set creation async stack if it's available regardless
existing of current parent async stack,
- we should cleanup parent link iff there is no creation and schedule async
stack for parent.

BUG=v8:6189
R=dgozman@chromium.org
==========

to

==========
[inspector] better stacks for promises

- we should always set creation async stack if it's available regardless
existing of current parent async stack,
- we should cleanup parent link iff there is no creation and schedule async
stack for parent.

Let's consider example: Promise.resolve().then(x => x).then(x => x), there is
three promises which will call following instrumentation:
1) created #1 (Promise.resolve()) - collected stack #1
2) scheduled #1 - collected stack #2
3) created #2 with #1 as parent (first .then) - collected stack #3
4) created #3 with #2 as parent (first .then) - collected stack #4
5) started #2 - use stack #2 as scheduled
6) scheduled #2 - collected stack #6
7) finished #2
8) started #3 - use stack #6 as scheduled 
9) scheduled #3 - collected stack #7
10) finished #3

If we collect stacks between step 4 and 5, it's possible to collect scheduled
stack #2 but still have creation stack for #2 - stack #3 - so we always need to
add creation event if scheduled is collected.

If we collect stacks between created and scheduled we should not remove parent
link even if parent was not scheduled yet.

BUG=v8:6189
R=dgozman@chromium.org
==========

[kozy, 2017-04-27 18:35:31]

I added details to description and answered to comments, please take another
look.

https://codereview.chromium.org/2844753002/diff/1/src/inspector/v8-debugger.cc
File src/inspector/v8-debugger.cc (left):

https://codereview.chromium.org/2844753002/diff/1/src/inspector/v8-debugger.c...
src/inspector/v8-debugger.cc:930: if (asyncParent.lock() && itCreation !=
m_asyncTaskCreationStacks.end()) {
this change is essential of this CL.

https://codereview.chromium.org/2844753002/diff/1/src/inspector/v8-debugger.cc
File src/inspector/v8-debugger.cc (right):

https://codereview.chromium.org/2844753002/diff/1/src/inspector/v8-debugger.c...
src/inspector/v8-debugger.cc:1053: // scheduled task#2 - stack#5
On 2017/04/27 17:26:34, dgozman wrote:
> scheduled after started?

yes, it's how our promise instrumentation works and why on started we get
scheduled stack from parent.

https://codereview.chromium.org/2844753002/diff/1/test/inspector/debugger/pro...
File test/inspector/debugger/promise-chain-when-limit-hit-expected.txt (right):

https://codereview.chromium.org/2844753002/diff/1/test/inspector/debugger/pro...
test/inspector/debugger/promise-chain-when-limit-hit-expected.txt:119:
inspector.setMaxAsyncTaskStacks(6)
On 2017/04/27 17:26:35, dgozman wrote:
> Why this one is shorter than previous?

Stack traces were collected at different moment of time and parent stack was
collected.

[dgozman, 2017-04-27 22:37:22]

https://codereview.chromium.org/2844753002/diff/1/src/inspector/v8-stack-trac...
File src/inspector/v8-stack-trace-impl.cc (right):

https://codereview.chromium.org/2844753002/diff/1/src/inspector/v8-stack-trac...
src/inspector/v8-stack-trace-impl.cc:80: (asyncParent->description() !=
description || !asyncParent->isEmpty() ||
We should not return here, but just skip the parent, continuing with parent's
parent.

[kozy, 2017-04-28 16:13:30]

all done, please take another look.

https://codereview.chromium.org/2844753002/diff/1/src/inspector/v8-stack-trac...
File src/inspector/v8-stack-trace-impl.cc (right):

https://codereview.chromium.org/2844753002/diff/1/src/inspector/v8-stack-trac...
src/inspector/v8-stack-trace-impl.cc:80: (asyncParent->description() !=
description || !asyncParent->isEmpty() ||
On 2017/04/27 22:37:22, dgozman wrote:
> We should not return here, but just skip the parent, continuing with parent's
> parent.

Done.

[kozy, 2017-04-28 16:13:35]

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

[commit-bot: I haz the power, 2017-04-28 16:13:39]

Dry run: 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, 2017-04-28 16:47:11]

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

[commit-bot: I haz the power, 2017-04-28 16:47:12]

Dry run: This issue passed the CQ dry run.

[dgozman, 2017-04-28 20:59:03]

lgtm

[kozy, 2017-04-28 21:05:18]

The CQ bit was checked by kozyatinskiy@chromium.org

[commit-bot: I haz the power, 2017-04-28 21:05:26]

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, 2017-04-28 21:07:02]

CQ is committing da patch.

Bot data: {"patchset_id": 20001, "attempt_start_ts": 1493413518009050,
"parent_rev": "6408032e6164f7f1b2f9edab75d1b14e4cbb32e4", "commit_rev":
"f2bd913cd459854894447f5422a953d23fa4e7ba"}

[commit-bot: I haz the power, 2017-04-28 21:07:10]

Description was changed from

==========
[inspector] better stacks for promises

- we should always set creation async stack if it's available regardless
existing of current parent async stack,
- we should cleanup parent link iff there is no creation and schedule async
stack for parent.

Let's consider example: Promise.resolve().then(x => x).then(x => x), there is
three promises which will call following instrumentation:
1) created #1 (Promise.resolve()) - collected stack #1
2) scheduled #1 - collected stack #2
3) created #2 with #1 as parent (first .then) - collected stack #3
4) created #3 with #2 as parent (first .then) - collected stack #4
5) started #2 - use stack #2 as scheduled
6) scheduled #2 - collected stack #6
7) finished #2
8) started #3 - use stack #6 as scheduled 
9) scheduled #3 - collected stack #7
10) finished #3

If we collect stacks between step 4 and 5, it's possible to collect scheduled
stack #2 but still have creation stack for #2 - stack #3 - so we always need to
add creation event if scheduled is collected.

If we collect stacks between created and scheduled we should not remove parent
link even if parent was not scheduled yet.

BUG=v8:6189
R=dgozman@chromium.org
==========

to

==========
[inspector] better stacks for promises

- we should always set creation async stack if it's available regardless
existing of current parent async stack,
- we should cleanup parent link iff there is no creation and schedule async
stack for parent.

Let's consider example: Promise.resolve().then(x => x).then(x => x), there is
three promises which will call following instrumentation:
1) created #1 (Promise.resolve()) - collected stack #1
2) scheduled #1 - collected stack #2
3) created #2 with #1 as parent (first .then) - collected stack #3
4) created #3 with #2 as parent (first .then) - collected stack #4
5) started #2 - use stack #2 as scheduled
6) scheduled #2 - collected stack #6
7) finished #2
8) started #3 - use stack #6 as scheduled 
9) scheduled #3 - collected stack #7
10) finished #3

If we collect stacks between step 4 and 5, it's possible to collect scheduled
stack #2 but still have creation stack for #2 - stack #3 - so we always need to
add creation event if scheduled is collected.

If we collect stacks between created and scheduled we should not remove parent
link even if parent was not scheduled yet.

BUG=v8:6189
R=dgozman@chromium.org

Review-Url: https://codereview.chromium.org/2844753002
Cr-Commit-Position: refs/heads/master@{#44990}
Committed:
https://chromium.googlesource.com/v8/v8/+/f2bd913cd459854894447f5422a953d23fa...
==========

[commit-bot: I haz the power, 2017-04-28 21:07:11]

Committed patchset #2 (id:20001) as
https://chromium.googlesource.com/v8/v8/+/f2bd913cd459854894447f5422a953d23fa...