Fix runit.py script on windows.

Fix runit.py script on windows.

Windows does not execute python scripts directly so it must call
the python executable

R=agable@chromium.org, dnj@chromium.org, petermayo@chromium.org
BUG=None
TEST=local, gclient runhooks works properly under a normal shell.

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=294048

[David Yen, 2015-02-12 00:54:48]

On 2015/02/12 00:47:49, David Yen wrote:

This is a continuation of https://codereview.chromium.org/888113002.

I'm not sure how it worked before, but under windows runhooks did not work for
me because the runit.py script directly calls the arguments as a subprocess. I
could fix it by prepending "build/scripts/common/cros_chromite.py" with
"python", but I thought it would be more correct to simply fix runit.py to work
under windows. Does anyone have an opinion on this?

[dnj, 2015-02-12 00:55:21]

dnj@chromium.org changed reviewers:
+ stip@chromium.org

[dnj, 2015-02-12 00:55:22]

+stip@, who wrote most of runit.py.

https://codereview.chromium.org/914303007/diff/1/scripts/tools/runit.py
File scripts/tools/runit.py (right):

https://codereview.chromium.org/914303007/diff/1/scripts/tools/runit.py#newco...
scripts/tools/runit.py:89: if 'win' in sys.platform.lower():
The best way to check for Windows is probably
https://docs.python.org/2/library/platform.html#platform.system

Any reason not to just do this for all OSes?

[David Yen, 2015-02-12 17:17:40]

https://codereview.chromium.org/914303007/diff/1/scripts/tools/runit.py
File scripts/tools/runit.py (right):

https://codereview.chromium.org/914303007/diff/1/scripts/tools/runit.py#newco...
scripts/tools/runit.py:89: if 'win' in sys.platform.lower():
On 2015/02/12 00:55:22, dnj wrote:
> The best way to check for Windows is probably
> https://docs.python.org/2/library/platform.html#platform.system
> 
> Any reason not to just do this for all OSes?

Only to be less intrusive, you're right that is is probably better. Done.

[dnj, 2015-02-12 17:55:41]

lgtm

[David Yen, 2015-02-12 19:39:08]

The CQ bit was checked by dyen@chromium.org

[commit-bot: I haz the power, 2015-02-12 19:39:56]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/914303007/20001

[commit-bot: I haz the power, 2015-02-12 19:43:07]

Committed patchset #2 (id:20001) as
http://src.chromium.org/viewvc/chrome?view=rev&revision=294048

[jam, 2015-02-12 21:13:39]

A revert of this CL (patchset #2 id:20001) has been created in
https://codereview.chromium.org/923723002/ by jam@chromium.org.

The reason for reverting is: Causing telemetry failures on bots, i.e.
http://build.chromium.org/p/tryserver.chromium.win/builders/win_chromium_rel_....

[David Yen, 2015-02-12 21:24:41]

On 2015/02/12 21:13:39, jam wrote:
> A revert of this CL (patchset #2 id:20001) has been created in
> https://codereview.chromium.org/923723002/ by mailto:jam@chromium.org.
> 
> The reason for reverting is: Causing telemetry failures on bots, i.e.
>
http://build.chromium.org/p/tryserver.chromium.win/builders/win_chromium_rel_....

It looks like invocations of the script have worked around this problem by
prepending python themselves when they are running on multiple platforms. Other
places that only run in linux scripts just pass in the python .py file as the
first argument.

I think the proper fix would be to always prepend python executable within the
script, and if the invoker would rather specify their own path to python have it
by done through a flag (ie. --python /usr/local/bin/python2.6) which basically
overrides sys.executable in the script.

However, I'm not too familiar with some of the code and cannot make the
distinction between when the invoker is trying to find a valid python file on a
certain platform or trying to target a specific python directory for test
reasons. For example in chromium_commands.py we have this:

    if self._target_platform == 'win32':
      py26 = J('src', 'third_party', 'python_26', 'python_slave.exe')
      test_cmd = ['cmd', '/C'] + [py26, script] + args
    elif self._target_platform == 'darwin':
      test_cmd = ['python2.6', script] + args
    elif self._target_platform == 'linux2':
      # Run thru runtest.py on linux to launch virtual x server
      test_cmd = self.GetTestCommand('/usr/local/bin/python2.6',
                                     [script] + args)

Are those python paths intentional for test reasons? Or are they merely known
locations where python would exist?