Refactor DeviceUtils.Install and InstallSplitApk to use the same helper

Refactor DeviceUtils.Install and InstallSplitApk to use the same helper

Motivated by Install() getting a new permissions attribute without
InstallSplitApk() being updated (http://crrev.com/de468d72d9e)

There is also a sematic change:
- When a previous install is present and reinstall=False, we used to
  check if the install on-device was identical and no-op in this case.
  Now, we do an uninstall + install regardless of whether the on-device
  version is identical. This new behaviour is more correct since it
  guarantees that no stale data will be present when reinstall=False.

BUG=

Committed: https://crrev.com/307cf30fea5e791de2ebac395b1d9a154afc637b
Cr-Commit-Position: refs/heads/master@{#348419}

[agrieve, 2015-09-11 15:20:41]

agrieve@chromium.org changed reviewers:
+ jbudorick@chromium.org

[jbudorick, 2015-09-11 15:26:16]

lgtm

[agrieve, 2015-09-11 16:47:05]

The CQ bit was checked by agrieve@chromium.org

[commit-bot: I haz the power, 2015-09-11 16:47:30]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1337073002/1
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1337073002/1

[commit-bot: I haz the power, 2015-09-11 17:30:32]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2015-09-11 17:31:12]

Patchset 1 (id:??) landed as
https://crrev.com/307cf30fea5e791de2ebac395b1d9a154afc637b
Cr-Commit-Position: refs/heads/master@{#348419}

[Dmitry Skiba, 2015-09-14 21:36:38]

dskiba@google.com changed reviewers:
+ dskiba@google.com

[Dmitry Skiba, 2015-09-14 21:36:39]

With this change consecutive invocations of "adb_install_apk.py --apk Chrome.apk
--release" always reinstall the app, while previously app was reinstalled only
once (and redundant invocations were fast).

I found this because manual perf tests became very slow for me. It turned out
that PossibleAndroidBrowser.UpdateExecutableIfNeeded()  (in
android_browser_finder.py) called twice during perf tests, and it now always
reinstalls.

[jbudorick, 2015-09-14 22:11:45]

On 2015/09/14 at 21:36:39, dskiba wrote:
> With this change consecutive invocations of "adb_install_apk.py --apk
Chrome.apk --release" always reinstall the app, while previously app was
reinstalled only once (and redundant invocations were fast).
> 
> I found this because manual perf tests became very slow for me. It turned out
that PossibleAndroidBrowser.UpdateExecutableIfNeeded()  (in
android_browser_finder.py) called twice during perf tests, and it now always
reinstalls.

That would be the semantic change described in the commit message. How much
slower are your perf test runs?

[Dmitry Skiba, 2015-09-14 22:40:37]

On 2015/09/14 22:11:45, jbudorick wrote:
> On 2015/09/14 at 21:36:39, dskiba wrote:
> > With this change consecutive invocations of "adb_install_apk.py --apk
> Chrome.apk --release" always reinstall the app, while previously app was
> reinstalled only once (and redundant invocations were fast).
> > 
> > I found this because manual perf tests became very slow for me. It turned
out
> that PossibleAndroidBrowser.UpdateExecutableIfNeeded()  (in
> android_browser_finder.py) called twice during perf tests, and it now always
> reinstalls.
> 
> That would be the semantic change described in the commit message. How much
> slower are your perf test runs?

Oops.

Well, I manually run perf tests with "run_benchmark run
memory.memory_health_plan --pageset-repeat 1" and it calls
android_browser_finder.py:UpdateExecutableIfNeeded() twice, and each reinstall
iteration takes ~40 seconds.

[agrieve, 2015-09-15 00:48:06]

On 2015/09/14 22:40:37, Dmitry Skiba wrote:
> On 2015/09/14 22:11:45, jbudorick wrote:
> > On 2015/09/14 at 21:36:39, dskiba wrote:
> > > With this change consecutive invocations of "adb_install_apk.py --apk
> > Chrome.apk --release" always reinstall the app, while previously app was
> > reinstalled only once (and redundant invocations were fast).
> > > 
> > > I found this because manual perf tests became very slow for me. It turned
> out
> > that PossibleAndroidBrowser.UpdateExecutableIfNeeded()  (in
> > android_browser_finder.py) called twice during perf tests, and it now always
> > reinstalls.
> > 
> > That would be the semantic change described in the commit message. How much
> > slower are your perf test runs?
> 
> Oops.
> 
> Well, I manually run perf tests with "run_benchmark run
> memory.memory_health_plan --pageset-repeat 1" and it calls
> android_browser_finder.py:UpdateExecutableIfNeeded() twice, and each reinstall
> iteration takes ~40 seconds.

Can we change the second invocation to use reinstall=True?

[commit-bot: I haz the power, 2015-09-23 12:23:27]

Patchset 1 (id:??) landed as
https://crrev.com/307cf30fea5e791de2ebac395b1d9a154afc637b
Cr-Commit-Position: refs/heads/master@{#348419}