[Telemetry] Change the way host port for android_browser_backend is created

[Telemetry] Change the way host port for android_browser_backend is created

Based on
https://code.google.com/p/chromium/codesearch#chromium/src/build/android/pylib/ports.py&l=40

AllocateTestServerPort used a lock file for storing tried port &
create a range limit for ports. Telemetry doesn't need this scheme
for finding a local port for these reasons:

1) There is no reason the local port has to be in the range 10201 & 30000 (there is
a TODO to get rid of it here:
https://code.google.com/p/chromium/codesearch#chromium/src/build/android/pyli...)
2) When the tried_port saved in lock file reaches the limit
TEST_SERVER_PORT_LAST, it creates the assertion failure in crbug.com/495715
3) The lock file make it any less flaky. Implementation of
GetUnreservedAvailableLocalPort() is guaranteed to return an
unused port number at the time of request.
Telemetry has been using this way for allocating port
for desktop testing just fine.

BUG=495715
Committed: https://crrev.com/a0e2e45b82acfc5ea63ecc1cee12b0cc1e86e557
Cr-Commit-Position: refs/heads/master@{#341546}

[nednguyen, 2015-08-03 15:09:13]

nednguyen@google.com changed reviewers:
+ jbudorick@chromium.org, sullivan@chromium.org

[jbudorick, 2015-08-03 15:11:06]

"telemetry doesn't need this scheme for finding a local port" -- why?

[nednguyen, 2015-08-03 15:30:42]

On 2015/08/03 15:11:06, jbudorick wrote:
> "telemetry doesn't need this scheme for finding a local port" -- why?

1) I don't get why the local port has to be in the range 10201 & 30000 (there is
a TODO to get rid of it here:
https://code.google.com/p/chromium/codesearch#chromium/src/build/android/pyli...)
2) When the tried_port saved in lock file reaches the limit
TEST_SERVER_PORT_LAST, it creates the assertion failure in crbug.com/495715
3) I don't think the lock file make it any less flaky. Implementation of
GetUnreservedAvailableLocalPort() is guaranteed to return an unused port number
at the time of request. Telemetry has been using this way for allocating port
for desktop testing just fine.

[jbudorick, 2015-08-03 15:32:14]

On 2015/08/03 at 15:30:42, nednguyen wrote:
> On 2015/08/03 15:11:06, jbudorick wrote:
> > "telemetry doesn't need this scheme for finding a local port" -- why?
> 
> 1) I don't get why the local port has to be in the range 10201 & 30000 (there
is a TODO to get rid of it here:
https://code.google.com/p/chromium/codesearch#chromium/src/build/android/pyli...)
> 2) When the tried_port saved in lock file reaches the limit
TEST_SERVER_PORT_LAST, it creates the assertion failure in crbug.com/495715
> 3) I don't think the lock file make it any less flaky. Implementation of
GetUnreservedAvailableLocalPort() is guaranteed to return an unused port number
at the time of request. Telemetry has been using this way for allocating port
for desktop testing just fine.

sgtm. I'm not that familiar with our port allocation, I just wanted a bit more
explanation for this change.

lgtm

[nednguyen, 2015-08-03 15:33:54]

The CQ bit was checked by nednguyen@google.com

[commit-bot: I haz the power, 2015-08-03 15:34:17]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1263293003/1
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1263293003/1

[commit-bot: I haz the power, 2015-08-03 18:01:20]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2015-08-03 18:01:57]

Patchset 1 (id:??) landed as
https://crrev.com/a0e2e45b82acfc5ea63ecc1cee12b0cc1e86e557
Cr-Commit-Position: refs/heads/master@{#341546}