Make gypv8sh.py compatible with ninja on Windows.

Make gypv8sh.py compatible with ninja on Windows.

This is because python's subprocess.py excepts GetStdHandle(STD_INPUT_HANDLE) to return a valid handle and it seems ninja on Windows doesn't give one, causing the subprocess.call() to fail unless we request stdin to be redirected.

Also fix file handle lifetime in gypv8sh.py

When an exception is thrown, the file handle could still be open, causing an
exception while trying to remove it in the exception handler due to file locking
on Windows.

Note that the exception being generated is another problem but this exception in
the exception handler was masking the original error.

R=scr@chromium.org
BUG=
TEST=

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=137902

[M-A Ruel, 2012-05-18 15:13:38]

Updated CL subject, cc'ed Scott so he is aware of the problem, to reproduce,
simply "ninja unit_tests"

[M-A Ruel, 2012-05-18 15:17:05]

FTR, here's the error message:

...\chrome\src\out\Release>ninja unit_tests
[2/650] RULE unit_tests: js2unit brows...w/print_preview_utils_unittest.gtestjs
FAILED: cmd /s /c "call "..\..\third_party\cygwin\setup_env.bat" && set
CYGWIN=nontsec && bash -c "cd ../../chrome ; 'python' '../tools/gypv8sh.py'
'../out/Release/v8_shell.exe' '../chrome/third_party/mock4js/mock4js.js'
'../chrome/test/data/webui/test_api.js' '../chrome/test/base/js2gtest.js' 'unit'
'browser/resources/print_preview/print_preview_utils_unittest.gtestjs'
'chrome/browser/resources/print_preview/print_preview_utils_unittest.gtestjs'
'../out/Release/obj/chrome/unit_tests.gen/chrome/browser/resources/print_preview/print_preview_utils_unittest-gen.cc'
'../out/Release/test_data/chrome/browser/resources/print_preview/print_preview_utils_unittest.gtestjs'""
Traceback (most recent call last):
  File "../tools/gypv8sh.py", line 55, in <module>
    sys.exit(main())
  File "../tools/gypv8sh.py", line 44, in main
    subprocess.check_call(cmd, stdout=f)
  File "...\chrome\src\third_party\python_26\lib\subprocess.py", line 457, in
check_call
    retcode = call(*popenargs, **kwargs)
  File "...\chrome\src\third_party\python_26\lib\subprocess.py", line 444, in
call
    return Popen(*popenargs, **kwargs).wait()
  File "...\chrome\src\third_party\python_26\lib\subprocess.py", line 588, in
__init__
    errread, errwrite) = self._get_handles(stdin, stdout, stderr)
  File "...\chrome\src\third_party\python_26\lib\subprocess.py", line 686, in
_get_handles
    p2cread = GetStdHandle(STD_INPUT_HANDLE)
WindowsError: [Error 6] The handle is invalid

[scottmg, 2012-05-18 15:40:21]

On 2012/05/18 15:13:38, Marc-Antoine Ruel wrote:
> Updated CL subject, cc'ed Scott so he is aware of the problem, to reproduce,
> simply "ninja unit_tests"

Thanks. I debugged and posted a similar change here:
https://chromiumcodereview.appspot.com/10388191/

I actually changed ninja to provide a valid stdin (an open handle to NUL,
instead of a NULL handle), but maybe I forgot to upload those binaries... sorry
about that. It seems like a good idea to make this script a bit more tolerant
anyway.

[scottmg, 2012-05-18 15:41:23]

> a similar change here:

Oops, wrong link: https://chromiumcodereview.appspot.com/10332081/

[M-A Ruel, 2012-05-18 15:43:42]

On 2012/05/18 15:40:21, scottmg wrote:
> On 2012/05/18 15:13:38, Marc-Antoine Ruel wrote:
> > Updated CL subject, cc'ed Scott so he is aware of the problem, to reproduce,
> > simply "ninja unit_tests"
> 
> Thanks. I debugged and posted a similar change here:
> https://chromiumcodereview.appspot.com/10388191/

Ok, but I think mine is better because it also fixes problem in the way the
error is reported. Any opinion?

[M-A Ruel, 2012-05-18 17:20:29]

Scott, want to approve? Sheridan is OOO today.

[scottmg, 2012-05-18 17:25:06]

lgtm