[Android] Expose provisioning settings as command line options

[Android] Expose provisioning settings as command line options

provision_devices.py currently relies on the builder name to determine
the desired device configuration for perf bots. To safely phase this out
this CL:
- exposes these configuration settings as command line options
- sets default values for these options based on the builder name, thus
  matching the previous behaviour of the script on all bots.
- explicitly adding these options on the perf bots will also maintain the
  same script behaviour, namely:
    --wait-for-battery --disable-network --disable-java-debug
- after all perf-bot recipes are updated, we can remove the
  builder-name-sniffing code.

In the process we also upgrade from optparse to argparse for handling
command line options.

BUG=450984
Committed: https://crrev.com/b205693d72e37d173f7dabcefe6c02f115509f27
Cr-Commit-Position: refs/heads/master@{#312818}

[perezju, 2015-01-22 12:07:12]

perezju@chromium.org changed reviewers:
+ jbudorick@chromium.org, klundberg@chromium.org

[perezju, 2015-01-22 12:07:12]

John, Karin: this CL would allow to configure the required settings for the bot
running the memory acceptance test, while not disturbing anything on the current
perf bots.

I would need some pointers later on how to update the bot recipes so that,
afterwards, we can safely remove the dependencies on the builder name.

https://codereview.chromium.org/869563002/diff/1/build/android/provision_devi...
File build/android/provision_devices.py (right):

https://codereview.chromium.org/869563002/diff/1/build/android/provision_devi...
build/android/provision_devices.py:100: ['chmod', '644',
constants.DEVICE_LOCAL_PROPERTIES_PATH],
apologies for the unrelated mending... couldn't resist :-/

https://codereview.chromium.org/869563002/diff/1/build/android/provision_devi...
build/android/provision_devices.py:143: device.Reboot(True, timeout=timeout,
retries=0)
On a occam-svelte device, reboot after wipe takes about 15 mins, quite some time
over the (now default value) of 10 mins to wait. Also thought that this would
allow to tune the value of the timeout, at least, on each different bot.

[perezju, 2015-01-22 12:09:04]

Also, not sure which would be suitable trybots for testing this before trying to
commit.

[jbudorick, 2015-01-22 14:20:28]

lgtm w/ nit

Sanity! Hooray!

https://codereview.chromium.org/869563002/diff/1/build/android/provision_devi...
File build/android/provision_devices.py (right):

https://codereview.chromium.org/869563002/diff/1/build/android/provision_devi...
build/android/provision_devices.py:100: ['chmod', '644',
constants.DEVICE_LOCAL_PROPERTIES_PATH],
On 2015/01/22 12:07:12, perezju wrote:
> apologies for the unrelated mending... couldn't resist :-/

This one seems pretty low risk. :)

https://codereview.chromium.org/869563002/diff/1/build/android/provision_devi...
build/android/provision_devices.py:143: device.Reboot(True, timeout=timeout,
retries=0)
On 2015/01/22 12:07:12, perezju wrote:
> On a occam-svelte device, reboot after wipe takes about 15 mins, quite some
time
> over the (now default value) of 10 mins to wait. Also thought that this would
> allow to tune the value of the timeout, at least, on each different bot.

This seems like a good idea to me. When I added 10 minute timeouts, I didn't
think any of the bots would come close...

https://codereview.chromium.org/869563002/diff/1/build/android/provision_devi...
build/android/provision_devices.py:277:
parser.add_argument('--disable-java-debug', action='store_true',
nit: I would probably have this use

  dest='enable_java_debug', default=True, action='store_false'

It may be easier to reason about in code that way, and it lines up better with
the java_debug parameter in _ConfigureLocalProperties.

[jbudorick, 2015-01-22 14:22:24]

On 2015/01/22 12:09:04, perezju wrote:
> Also, not sure which would be suitable trybots for testing this before trying
to
> commit.

android_dbg_tests_recipe

I don't know of any perf trybots.

[klundberg, 2015-01-22 15:30:04]

On 2015/01/22 14:22:24, jbudorick wrote:
> On 2015/01/22 12:09:04, perezju wrote:
> > Also, not sure which would be suitable trybots for testing this before
trying
> to
> > commit.
> 
> android_dbg_tests_recipe
> 
> I don't know of any perf trybots.

There are indeed no perf trybots. 

The closest is android_rel_tests_recipe which runs the Telemetry unit tests but
they are more or less consistently red so that won't help you.

[perezju, 2015-01-23 10:54:32]

https://codereview.chromium.org/869563002/diff/1/build/android/provision_devi...
File build/android/provision_devices.py (right):

https://codereview.chromium.org/869563002/diff/1/build/android/provision_devi...
build/android/provision_devices.py:277:
parser.add_argument('--disable-java-debug', action='store_true',
On 2015/01/22 14:20:28, jbudorick wrote:
> nit: I would probably have this use
> 
>   dest='enable_java_debug', default=True, action='store_false'
> 
> It may be easier to reason about in code that way, and it lines up better with
> the java_debug parameter in _ConfigureLocalProperties.

Yeah, I went back and forth with this idea as well, and wasn't quite sure which
way it was easier to reason about. Anyway, I updated the code and went with the
approach you suggested.

[perezju, 2015-01-23 10:54:53]

The CQ bit was checked by perezju@chromium.org

[commit-bot: I haz the power, 2015-01-23 10:55:36]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/869563002/20001

[commit-bot: I haz the power, 2015-01-23 11:56:11]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-01-23 11:56:57]

Patchset 2 (id:??) landed as
https://crrev.com/b205693d72e37d173f7dabcefe6c02f115509f27
Cr-Commit-Position: refs/heads/master@{#312818}