Make timeline model able to return browser process

Make timeline model able to return browser process

When parsing tracing data into timeline model, also find the browser
process and save it into timeline model. This is to support querying
async slices starting in browser process.

BUG=246034
TEST=Run telemetry smoothness test and make sure input latency trace
     can be retrieved from querying async slices on browser process.

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=247911

[Yufeng Shen (Slow to review), 2014-01-23 22:06:14]

This is forked from 
https://codereview.chromium.org/132433004/  as Nat suggested.


While writing the CL, I found the easy way to get the browser process
would be just call timeline.GetAllProcesses() and find the one with
name 'Browser'.  But as Nat said, to make the timeline model browser
agnostic, it is better to isolate this kind of browser-specific string
query into chrome_trace_result.

[nduca, 2014-01-28 00:02:43]

looks good, but needs tests on all added features please.

https://codereview.chromium.org/144423003/diff/1/tools/telemetry/telemetry/co...
File tools/telemetry/telemetry/core/backends/chrome/chrome_trace_result.py
(right):

https://codereview.chromium.org/144423003/diff/1/tools/telemetry/telemetry/co...
tools/telemetry/telemetry/core/backends/chrome/chrome_trace_result.py:24: for
process in timeline.GetAllProcesses():
Please write a unit test that validates that chrome traces have a browser
process set.

[epennerAtGoogle, 2014-01-28 00:30:34]

On 2014/01/28 00:02:43, nduca wrote:
> looks good, but needs tests on all added features please.
> 
>
https://codereview.chromium.org/144423003/diff/1/tools/telemetry/telemetry/co...
> File tools/telemetry/telemetry/core/backends/chrome/chrome_trace_result.py
> (right):
> 
>
https://codereview.chromium.org/144423003/diff/1/tools/telemetry/telemetry/co...
> tools/telemetry/telemetry/core/backends/chrome/chrome_trace_result.py:24: for
> process in timeline.GetAllProcesses():
> Please write a unit test that validates that chrome traces have a browser
> process set.

Looking good. However I meant to comment on this concept before so I'll just
throw it out there. The approach that's being taken is to abstract "Render
Process" and "Browser Process" as much as possible. That's good if we need those
things, but I think another approach would be to never assume any model has this
or that process... The troubles begin when the GPU process is needed (doesn't
always exist) or we are in webview and there is no renderer or GPU process.

In most cases I think process threads can just be 'merged' for analysis
purposes, and the only assumption then becomes thread naming conventions.

In the few cases where processes are needed I think it would likely be to
distinguish renderer threads in a multi-tab use case?

[nduca, 2014-01-28 06:53:33]

In @miletus' case, he needs to find the browser process, where we have latency
event records.

One option might be, I suppose, to just iterate all the events everywhere by
name looking for the latency event records...

[epennerAtGoogle, 2014-01-28 19:55:56]

On 2014/01/28 06:53:33, nduca wrote:
> In @miletus' case, he needs to find the browser process, where we have latency
> event records.
> 
> One option might be, I suppose, to just iterate all the events everywhere by
> name looking for the latency event records...

Regarding the latency events, are those tied to a thread? CrBrowserMain thread
is always available and unique in the list of all threads, no matter the process
configuration.

Hmm, however, the browser process is actually one thing where we will always
have one (I hope!), so maybe that's totally fine. We should just be careful of
things that assume there is renderer processes or GPU processes.

Btw, I'm landing a CL to label the browser proc on Android.

[Yufeng Shen (Slow to review), 2014-01-29 20:33:20]

PTAL, thanks!

https://codereview.chromium.org/144423003/diff/1/tools/telemetry/telemetry/co...
File tools/telemetry/telemetry/core/backends/chrome/chrome_trace_result.py
(right):

https://codereview.chromium.org/144423003/diff/1/tools/telemetry/telemetry/co...
tools/telemetry/telemetry/core/backends/chrome/chrome_trace_result.py:24: for
process in timeline.GetAllProcesses():
On 2014/01/28 00:02:43, nduca wrote:
> Please write a unit test that validates that chrome traces have a browser
> process set.

unittest added.

[nduca, 2014-01-29 21:03:41]

lgtm with nit

https://codereview.chromium.org/144423003/diff/40001/tools/telemetry/telemetr...
File tools/telemetry/telemetry/core/backends/chrome/chrome_trace_result.py
(right):

https://codereview.chromium.org/144423003/diff/40001/tools/telemetry/telemetr...
tools/telemetry/telemetry/core/backends/chrome/chrome_trace_result.py:24: for
process in timeline.GetAllProcesses():
can you find threads named CrRendererMain and then take its parent to get
browser process? It will work with older chromium builds... telemetry tries when
possible to work on old builds and as we noted previously, the name 'Browser'
works on android only in the last few days.

https://codereview.chromium.org/144423003/diff/40001/tools/telemetry/telemetr...
File tools/telemetry/telemetry/core/backends/chrome/tracing_backend_unittest.py
(right):

https://codereview.chromium.org/144423003/diff/40001/tools/telemetry/telemetr...
tools/telemetry/telemetry/core/backends/chrome/tracing_backend_unittest.py:137:
def testBrowserProcess(self):
unit test for the old way too please, e.g. a thread named CrBrowserMain

[Yufeng Shen (Slow to review), 2014-01-29 21:23:51]

https://codereview.chromium.org/144423003/diff/40001/tools/telemetry/telemetr...
File tools/telemetry/telemetry/core/backends/chrome/chrome_trace_result.py
(right):

https://codereview.chromium.org/144423003/diff/40001/tools/telemetry/telemetr...
tools/telemetry/telemetry/core/backends/chrome/chrome_trace_result.py:24: for
process in timeline.GetAllProcesses():
On 2014/01/29 21:03:41, nduca wrote:
> can you find threads named CrRendererMain and then take its parent to get
> browser process? It will work with older chromium builds... telemetry tries
when
> possible to work on old builds and as we noted previously, the name 'Browser'
> works on android only in the last few days.

Done.

https://codereview.chromium.org/144423003/diff/40001/tools/telemetry/telemetr...
File tools/telemetry/telemetry/core/backends/chrome/tracing_backend_unittest.py
(right):

https://codereview.chromium.org/144423003/diff/40001/tools/telemetry/telemetr...
tools/telemetry/telemetry/core/backends/chrome/tracing_backend_unittest.py:137:
def testBrowserProcess(self):
On 2014/01/29 21:03:41, nduca wrote:
> unit test for the old way too please, e.g. a thread named CrBrowserMain

Done.

[commit-bot: I haz the power, 2014-01-29 21:37:35]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/miletus@chromium.org/144423003/60001

[commit-bot: I haz the power, 2014-01-30 03:04:19]

CQ bit was unchecked on CL. Ignoring.

[commit-bot: I haz the power, 2014-01-30 03:11:13]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/miletus@chromium.org/144423003/60001

[commit-bot: I haz the power, 2014-01-30 09:10:51]

Change committed as 247911