Android: Make netlog available in startup traces.

Android: Make netlog available in startup traces.

Netlog events are made available to trace events through
TraceNetLogObserver. On Android, it is created after tracing has
started, yet only listens for tracing state changes. Before listening
for tracing state changes, first check whether tracing is already
enabled.

TEST=
$ adb shell 'echo "_ --trace-startup=*,netlog" > /data/local/tmp/chrome-command-line'
$ adb shell 'am force-stop com.google.android.apps.chrome' &&
  adb shell am start -a 'android.intent.action.VIEW' -n com.google.android.apps.chrome/.Main -d 'https://android.com'"

This was busted before, works with this patch.

BUG=668195
Committed: https://crrev.com/9da0f1a6adf25390c93d0ddd384defcedda25f89
Cr-Commit-Position: refs/heads/master@{#435895}

[Benoit L, 2016-11-23 17:09:15]

[email protected] changed reviewers:
+ [email protected]

[Benoit L, 2016-11-23 17:09:21]

The CQ bit was checked by [email protected] to run a CQ dry run

[commit-bot: I haz the power, 2016-11-23 17:09:44]

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, 2016-11-23 17:45:31]

The CQ bit was unchecked by [email protected]

[commit-bot: I haz the power, 2016-11-23 17:45:32]

Dry run: Try jobs failed on following builders:
  ios-simulator on master.tryserver.chromium.mac (JOB_FAILED,
http://build.chromium.org/p/tryserver.chromium.mac/builders/ios-simulator/bui...)

[eroman, 2016-11-23 20:36:01]

[email protected] changed reviewers:
+ [email protected]

[eroman, 2016-11-23 20:36:01]

+Helen who is more familiar with the tracing integration.

(1) How does this relate to the existing test
TraceNetLogObserverTest.CreateObserverAfterTracingStarts ? Looks like that test
explicitly was not handling this case, would like to understand from Helen if
there is a concern with changing this policy.

(2) A consequence of this change is that we will end up calling a virtual
function within the constructor, of the object being constructed (i.e. while
inside ChromeNetLog::ChromeNetLog we will now indireclty call
this->DeprecatedAddObserver()). This is generally a bad pattern as it breaks on
base classes. It *should* work in this case since nothing extends
ChromeNetLog... Can you structure the code differently to avoid this?

[xunjieli, 2016-11-29 14:45:44]

On 2016/11/23 20:36:01, eroman (slow) wrote:
> +Helen who is more familiar with the tracing integration.
> 
> (1) How does this relate to the existing test
> TraceNetLogObserverTest.CreateObserverAfterTracingStarts ? Looks like that
test
> explicitly was not handling this case, would like to understand from Helen if
> there is a concern with changing this policy.

I think it is fine to change the policy here. Could you add an explicit check to
make sure NetLog category is selected in startup tracing? I commented in the
bug. My main concern is that NetLog trace events can take up a lot of space.
Since we don't have the trace viewer UI and the prompt, I think we should
disable it by default.
 
> 
> (2) A consequence of this change is that we will end up calling a virtual
> function within the constructor, of the object being constructed (i.e. while
> inside ChromeNetLog::ChromeNetLog we will now indireclty call
> this->DeprecatedAddObserver()). This is generally a bad pattern as it breaks
on
> base classes. It *should* work in this case since nothing extends
> ChromeNetLog... Can you structure the code differently to avoid this?

[Benoit L, 2016-11-30 16:58:30]

On 2016/11/29 14:45:44, xunjieli wrote:
> On 2016/11/23 20:36:01, eroman (slow) wrote:
> > +Helen who is more familiar with the tracing integration.
> > 
> > (1) How does this relate to the existing test
> > TraceNetLogObserverTest.CreateObserverAfterTracingStarts ? Looks like that
> test
> > explicitly was not handling this case, would like to understand from Helen
if
> > there is a concern with changing this policy.
> 
> I think it is fine to change the policy here. Could you add an explicit check
to
> make sure NetLog category is selected in startup tracing? I commented in the
> bug. My main concern is that NetLog trace events can take up a lot of space.
> Since we don't have the trace viewer UI and the prompt, I think we should
> disable it by default.
>  
> > 
> > (2) A consequence of this change is that we will end up calling a virtual
> > function within the constructor, of the object being constructed (i.e. while
> > inside ChromeNetLog::ChromeNetLog we will now indireclty call
> > this->DeprecatedAddObserver()). This is generally a bad pattern as it breaks
> on
> > base classes. It *should* work in this case since nothing extends
> > ChromeNetLog... Can you structure the code differently to avoid this?

Per the discussion in the bug:
- netlog is enabled as usual, no special casing for startup tracing.
- Added a test to check that if the category is disabled, nothing gets recorded.
This one may not be necessary, as this is no different from anything else in
tracing.

Note that a trace configuration can be provided for startup tracing, either
through command-line flags or through the trace config file. Since changing the
flags is necessary to enable startup tracing anyway, then the control does exist
for users.

About the virtual function call: NetLog::DeprecatedAddObserver() is not virtual,
and NetLog has no virtual member function (aside from the destructor). So it
should be fine, right?

[Benoit L, 2016-11-30 16:58:48]

The CQ bit was checked by [email protected] to run a CQ dry run

[commit-bot: I haz the power, 2016-11-30 16:59:33]

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, 2016-11-30 17:53:09]

The CQ bit was unchecked by [email protected]

[commit-bot: I haz the power, 2016-11-30 17:53:09]

Dry run: This issue passed the CQ dry run.

[xunjieli, 2016-11-30 18:12:32]

Acknowledged. LGTM. Please wait for eroman@ to sign off.

https://codereview.chromium.org/2524043002/diff/20001/net/log/trace_net_log_o...
File net/log/trace_net_log_observer_unittest.cc (right):

https://codereview.chromium.org/2524043002/diff/20001/net/log/trace_net_log_o...
net/log/trace_net_log_observer_unittest.cc:327:
set_trace_net_log_observer(NULL);
nit: use nullptr instead of NULL in new code.

[Benoit L, 2016-12-01 10:41:32]

Thanks!

eroman@: PTAL, thank you.

https://codereview.chromium.org/2524043002/diff/20001/net/log/trace_net_log_o...
File net/log/trace_net_log_observer_unittest.cc (right):

https://codereview.chromium.org/2524043002/diff/20001/net/log/trace_net_log_o...
net/log/trace_net_log_observer_unittest.cc:327:
set_trace_net_log_observer(NULL);
On 2016/11/30 18:12:32, xunjieli wrote:
> nit: use nullptr instead of NULL in new code.

Done.

[Benoit L, 2016-12-01 16:26:45]

The CQ bit was checked by [email protected] to run a CQ dry run

[commit-bot: I haz the power, 2016-12-01 16:27:05]

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, 2016-12-01 17:32:18]

The CQ bit was unchecked by [email protected]

[commit-bot: I haz the power, 2016-12-01 17:32:19]

Dry run: This issue passed the CQ dry run.

[eroman, 2016-12-01 20:42:13]

lgtm

[Benoit L, 2016-12-02 09:22:16]

The CQ bit was checked by [email protected]

[Benoit L, 2016-12-02 09:22:16]

The patchset sent to the CQ was uploaded after l-g-t-m from
[email protected]
Link to the patchset: https://codereview.chromium.org/2524043002/#ps40001
(title: "Address comment.")

[commit-bot: I haz the power, 2016-12-02 09:22:40]

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, 2016-12-02 09:26:51]

CQ is committing da patch.

Bot data: {"patchset_id": 40001, "attempt_start_ts": 1480670536264580,
"parent_rev": "8d92448b226699e67088c17ece7660bf71d9f7e0", "commit_rev":
"69520e245db98ef4a719e20018ce29ea6844fb1c"}

[commit-bot: I haz the power, 2016-12-02 09:27:19]

Committed patchset #3 (id:40001)

[commit-bot: I haz the power, 2016-12-02 09:30:24]

Description was changed from

==========
Android: Make netlog available in startup traces.

Netlog events are made available to trace events through
TraceNetLogObserver. On Android, it is created after tracing has
started, yet only listens for tracing state changes. Before listening
for tracing state changes, first check whether tracing is already
enabled.

TEST=
$ adb shell 'echo "_ --trace-startup=*,netlog" >
/data/local/tmp/chrome-command-line'
$ adb shell 'am force-stop com.google.android.apps.chrome' &&
  adb shell am start -a 'android.intent.action.VIEW' -n
com.google.android.apps.chrome/.Main -d 'https://android.com'"

This was busted before, works with this patch.

BUG=668195
==========

to

==========
Android: Make netlog available in startup traces.

Netlog events are made available to trace events through
TraceNetLogObserver. On Android, it is created after tracing has
started, yet only listens for tracing state changes. Before listening
for tracing state changes, first check whether tracing is already
enabled.

TEST=
$ adb shell 'echo "_ --trace-startup=*,netlog" >
/data/local/tmp/chrome-command-line'
$ adb shell 'am force-stop com.google.android.apps.chrome' &&
  adb shell am start -a 'android.intent.action.VIEW' -n
com.google.android.apps.chrome/.Main -d 'https://android.com'"

This was busted before, works with this patch.

BUG=668195

Committed: https://crrev.com/9da0f1a6adf25390c93d0ddd384defcedda25f89
Cr-Commit-Position: refs/heads/master@{#435895}
==========

[commit-bot: I haz the power, 2016-12-02 09:30:25]

Patchset 3 (id:??) landed as
https://crrev.com/9da0f1a6adf25390c93d0ddd384defcedda25f89
Cr-Commit-Position: refs/heads/master@{#435895}