android: Post stop/onServiceDisconnected to launcher

android: Post stop/onServiceDisconnected to launcher

From native, stop used to be called on both the client (UI/IO) thread
and on launcher thread, depending only which ChildProcessLauncher
method client calls. Now post both to launcher thread.

onServiceDisconnected to launcher thread as well. onServiceDisconnected
also calls stop, so these two changes are bundled together.

This allows all synchronization to be removed from Allocator since
everything is accessed on the Launcher thread only. But once again,
tests needed to be fixed here as well.

Note onServiceConnected is unchanged because it has some interaction
with the warm up connection.

Potential risks:
I think delaying native calls are fine. stop was always asynchronous
anyway. There is a bit of a risk delaying onServiceDisconnected because
dropping bindings will prevent android from restarting the service on
our behalf. I think we just have to swallow this latency and one day
unify everything to run on the launcher thread.
Of course there could be other unforeseen issues.

BUG=689758
Review-Url: https://codereview.chromium.org/2821583002
Cr-Commit-Position: refs/heads/master@{#465139}
Committed: https://chromium.googlesource.com/chromium/src/+/da4f4f98353bc5fd8f88300721260355bb994b3b

[boliu, 2017-04-13 22:32:49]

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

[commit-bot: I haz the power, 2017-04-13 22:33:31]

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, 2017-04-13 23:14:42]

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

[commit-bot: I haz the power, 2017-04-13 23:14:43]

Dry run: This issue passed the CQ dry run.

[boliu, 2017-04-13 23:18:07]

boliu@chromium.org changed reviewers:
+ mariakhomenko@chromium.org

[boliu, 2017-04-13 23:18:07]

ptal. again maria to review, jay fyi

I think I'll wait until tomorrow after branch to land this. Want to leave plenty
of time in case this does break something..

[boliu, 2017-04-14 18:31:31]

Description was changed from

==========
android: Post stop/onServiceDisconnected to launcher

From native, stop used to be called on both the client (UI/IO) thread
and on launcher thread, depending only which ChildProcessLauncher
method client calls. Now post both to launcher thread.

onServiceDisconnected to launcher thread as well. onServiceDisconnected
also calls stop, so these two changes are bundled together.

Note onServiceConnected is unchanged because it has some interaction
with the warm up connection.

Potential risks:
I think delaying native calls are fine. stop was always asynchronous
anyway. There is a bit of a risk delaying onServiceDisconnected because
dropping bindings will prevent android from restarting the service on
our behalf. I think we just have to swallow this latency and one day
unify everything to run on the launcher thread.
Of course there could be other unforeseen issues.

BUG=689758
==========

to

==========
android: Post stop/onServiceDisconnected to launcher

From native, stop used to be called on both the client (UI/IO) thread
and on launcher thread, depending only which ChildProcessLauncher
method client calls. Now post both to launcher thread.

onServiceDisconnected to launcher thread as well. onServiceDisconnected
also calls stop, so these two changes are bundled together.

This allows all synchronization to be removed from Allocator since
everything is accessed on the Launcher thread only. But once again,
tests needed to be fixed here as well.

Note onServiceConnected is unchanged because it has some interaction
with the warm up connection.

Potential risks:
I think delaying native calls are fine. stop was always asynchronous
anyway. There is a bit of a risk delaying onServiceDisconnected because
dropping bindings will prevent android from restarting the service on
our behalf. I think we just have to swallow this latency and one day
unify everything to run on the launcher thread.
Of course there could be other unforeseen issues.

BUG=689758
==========

[boliu, 2017-04-14 18:32:23]

maria is ooo today, so I'm just piling more things onto this CL. turns out with
one more change, this allows all the synchronization to be removed from
Allocator \o/ so did that

[boliu, 2017-04-14 18:32:28]

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

[commit-bot: I haz the power, 2017-04-14 18:32:56]

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, 2017-04-14 19:11:25]

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

[commit-bot: I haz the power, 2017-04-14 19:11:26]

Dry run: This issue passed the CQ dry run.

[Maria, 2017-04-18 03:48:41]

lgtm

[boliu, 2017-04-18 03:56:29]

The CQ bit was checked by boliu@chromium.org

[commit-bot: I haz the power, 2017-04-18 03:56:46]

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, 2017-04-18 04:40:24]

CQ is committing da patch.

Bot data: {"patchset_id": 40001, "attempt_start_ts": 1492487789409610,
"parent_rev": "196a305434d54e05c519f1f26e129d7cf2f863ba", "commit_rev":
"da4f4f98353bc5fd8f88300721260355bb994b3b"}

[commit-bot: I haz the power, 2017-04-18 04:40:32]

Description was changed from

==========
android: Post stop/onServiceDisconnected to launcher

From native, stop used to be called on both the client (UI/IO) thread
and on launcher thread, depending only which ChildProcessLauncher
method client calls. Now post both to launcher thread.

onServiceDisconnected to launcher thread as well. onServiceDisconnected
also calls stop, so these two changes are bundled together.

This allows all synchronization to be removed from Allocator since
everything is accessed on the Launcher thread only. But once again,
tests needed to be fixed here as well.

Note onServiceConnected is unchanged because it has some interaction
with the warm up connection.

Potential risks:
I think delaying native calls are fine. stop was always asynchronous
anyway. There is a bit of a risk delaying onServiceDisconnected because
dropping bindings will prevent android from restarting the service on
our behalf. I think we just have to swallow this latency and one day
unify everything to run on the launcher thread.
Of course there could be other unforeseen issues.

BUG=689758
==========

to

==========
android: Post stop/onServiceDisconnected to launcher

From native, stop used to be called on both the client (UI/IO) thread
and on launcher thread, depending only which ChildProcessLauncher
method client calls. Now post both to launcher thread.

onServiceDisconnected to launcher thread as well. onServiceDisconnected
also calls stop, so these two changes are bundled together.

This allows all synchronization to be removed from Allocator since
everything is accessed on the Launcher thread only. But once again,
tests needed to be fixed here as well.

Note onServiceConnected is unchanged because it has some interaction
with the warm up connection.

Potential risks:
I think delaying native calls are fine. stop was always asynchronous
anyway. There is a bit of a risk delaying onServiceDisconnected because
dropping bindings will prevent android from restarting the service on
our behalf. I think we just have to swallow this latency and one day
unify everything to run on the launcher thread.
Of course there could be other unforeseen issues.

BUG=689758

Review-Url: https://codereview.chromium.org/2821583002
Cr-Commit-Position: refs/heads/master@{#465139}
Committed:
https://chromium.googlesource.com/chromium/src/+/da4f4f98353bc5fd8f8830072126...
==========

[commit-bot: I haz the power, 2017-04-18 04:40:32]

Committed patchset #3 (id:40001) as
https://chromium.googlesource.com/chromium/src/+/da4f4f98353bc5fd8f8830072126...