IPC: Make ipc_perftests run on Android.

IPC: Make ipc_perftests run on Android.

There was several minor issues:

- Base perf logging file location was off-limits on Android

- Printf needs to be flushed to be visible on Android.

- Android needs to reset the 'PipeMap' manually since we can't 'exec' after 
forking a test process. If we don't do this the Channel thinks we are in a single-
process test and tries to open an FD which was closed during forking.

- Android's base file descriptor needs to be increased to prevent stomping 
the android native logging file-descriptor with the default pipe.

- The test took too long, so the 'exponent' is reduced from 5 to 3

- We need an APK

With this patch the test runs like on other platforms, and lots of testing code is
fixed such that it works the same way on all platforms.

BUG=345471
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=257877

[epenner, 2014-03-15 02:19:33]

Ptal.

Needed to learn quite a bit about our multi-process tests and channels etc. to
figure this out. Works just fine now and several little bits of testing code now
work on Android.

[Nico, 2014-03-15 03:06:14]

base/ lgtm

https://codereview.chromium.org/196343019/diff/20001/ipc/ipc_channel.h
File ipc/ipc_channel.h (right):

https://codereview.chromium.org/196343019/diff/20001/ipc/ipc_channel.h#newcod...
ipc/ipc_channel.h:211: // process such that it act's similar to if it was
exec'd, for tests.
acts

[viettrungluu, 2014-03-15 17:25:27]

LGTM w/comments.

https://codereview.chromium.org/196343019/diff/20001/base/test/perf_log.cc
File base/test/perf_log.cc (right):

https://codereview.chromium.org/196343019/diff/20001/base/test/perf_log.cc#ne...
base/test/perf_log.cc:42: fflush(stdout);
I'm not sure why you fflush() before the printf().... <shrug>

https://codereview.chromium.org/196343019/diff/20001/ipc/ipc_perftests.cc
File ipc/ipc_perftests.cc (right):

https://codereview.chromium.org/196343019/diff/20001/ipc/ipc_perftests.cc#new...
ipc/ipc_perftests.cc:235: int msg_count = 100000;
You might consider lowering msg_count instead of lowering kMsgSizeMaxExp, or
making msg_count dependent on msg_size. (We mostly send small messages, but it's
also good to know the performance characteristics on bigger messages.)

(Currently, at max exp of 3, that's a msg_size of 1728. If that runs in an
acceptable amount of time, you can definitely send 1728 * 100000 bytes in an
acceptable amount of time. So msg_count = 1728 * 100000 / msg_size would be
plausible for larger messages. It might be too big for smaller messages.)

[epenner, 2014-03-17 20:24:53]

https://codereview.chromium.org/196343019/diff/20001/base/test/perf_log.cc
File base/test/perf_log.cc (right):

https://codereview.chromium.org/196343019/diff/20001/base/test/perf_log.cc#ne...
base/test/perf_log.cc:42: fflush(stdout);
On 2014/03/15 17:25:27, viettrungluu wrote:
> I'm not sure why you fflush() before the printf().... <shrug>

I must admit I just lifted this from some other perf-tests. I've removed the
first since indeed it seems redundant.

https://codereview.chromium.org/196343019/diff/20001/ipc/ipc_channel.h
File ipc/ipc_channel.h (right):

https://codereview.chromium.org/196343019/diff/20001/ipc/ipc_channel.h#newcod...
ipc/ipc_channel.h:211: // process such that it act's similar to if it was
exec'd, for tests.
On 2014/03/15 03:06:15, Nico (on GMT time Mar 15 - 24) wrote:
> acts

Done.

https://codereview.chromium.org/196343019/diff/20001/ipc/ipc_perftests.cc
File ipc/ipc_perftests.cc (right):

https://codereview.chromium.org/196343019/diff/20001/ipc/ipc_perftests.cc#new...
ipc/ipc_perftests.cc:235: int msg_count = 100000;
Good point.

Okay I kept all 5 message sizes. I simply put all the sizes/counts into arrays.
Let me know if there is any issues, otherwise I'm going to land.

If I add more tests here I'll make it dynamic (ie 'run for 3 seconds').

[epenner, 2014-03-17 20:30:37]

Actually, I still need OWNERS review for IPC.

The primary IPC change here is accommodating our inability to 'exec' on Android
(at least for simple tests).

[cpu_(ooo_6.6-7.5), 2014-03-18 17:19:50]

lgtm

[epenner, 2014-03-18 21:56:47]

The CQ bit was checked by epenner@chromium.org

[commit-bot: I haz the power, 2014-03-18 22:12:27]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/epenner@chromium.org/196343019/60001

[epenner, 2014-03-18 22:16:00]

The CQ bit was checked by epenner@chromium.org

[commit-bot: I haz the power, 2014-03-18 22:21:35]

Failed to trigger a try job on win_x64_rel
HTTP Error 400: Bad Request

[commit-bot: I haz the power, 2014-03-18 22:26:00]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/epenner@chromium.org/196343019/80001

[commit-bot: I haz the power, 2014-03-18 22:45:42]

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

[commit-bot: I haz the power, 2014-03-18 22:45:43]

Try jobs failed on following builders:
  tryserver.chromium on linux_chromium_chromeos_rel

[epenner, 2014-03-18 23:39:34]

The CQ bit was checked by epenner@chromium.org

[commit-bot: I haz the power, 2014-03-18 23:43:08]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/epenner@chromium.org/196343019/80001

[commit-bot: I haz the power, 2014-03-19 06:34:54]

Change committed as 257877