Python 3 compatibility

Well this ended up being bigger than I anticipated!
As well as being compatible with python 2, this changeset makes gyp (almost) compatible with python 3 as well.
This is done without introducing any new module dependencies.

As it's such a large changeset it may be easier to view it commit by commit, which you can do here: https://github.com/AWhetter/gyp/compare/python3Compat...python3Compatibility
If this needs splitting into more changesets to make it more manageable then let me know! I realise that this is a big change to land in one go.
Half of the line changes come just from updating print statements to include brackets!
It's maybe worth looking at the commits anyway because I've tried to make the commit messages describe my thinking behind the change.

I've tried to make the changes in here mainly syntax changes. I've put more "functional" changes into other changesets:
Replaced custom OrderedDict with collections.OrderedDict (https://codereview.chromium.org/1429683005)
Replaced deprecated os.path.walk usage with os.walk (https://codereview.chromium.org/1429933006)
Replaced deprecated compiler module usage with the ast module (https://codereview.chromium.org/1406163007)

There are some tests that still fail under python 3 but not under python 2.
The test/variables/commands tests fail because the order of iteration of the parsed dictionary is not guaranteed. In python 2 this never seemed to be an issue but in python 3 the iteration order does changes intermittently. This in turn causes the print order of debug messages to change, and therefore not always match against the expected output. The way this test is checked needs to be reworked to pass under python 3.
The other tests that fail under python 3 are the MSVS settings tests. This is a genuine test fail that is again caused by the order of dictionary iteration. On line 466 of pylib/gyp/MSVSSettings.py we iterate over the msvs_tool_settings items. The 'DefaultCharIsUnsigned' value requires 'AdditionalOptions' to have been converted first so that it resides in the msbuild_settings dictionary under the appropriate tool. A quick and dirty fix is to make msvs_tool_settings.items() sorted on this line.
I wasn't sure how best to go about resolving these issues so they're up for discussion.

Fixes issue #36 (https://bugs.chromium.org/p/gyp/issues/detail?id=36)

[AWhetter, 2015-11-16 21:41:16]

Description was changed from

==========
Python 3 compatibility

Made error tests more reliable

The order in which the nodes of a cycle are printed is unpredictable.
I believe this is because the order of iteration over the items in the
dependency dictionary is unpredictable.
Therefore the error tests now match against all the possible orderings of
dependencies in a detected cycle.

Fixed a cyclic dependency on python 3

Down the import dependency chain of copy.py the 'tarfile' module was trying to
import the 'copy' standard library module but was finding the copy.py file first
because it was in the current directory.

Made file reading and writing compatible with python 3

Mostly this is opening files in text mode rather than binary mode, so that
inputs and outputs are not 'bytes' and remain strings.
Where subprocess output is compared with file contents, strings, or type
checked, 'universal_newlines' has been turned on so that it also outputs a
string rather than a 'bytes'.

Fixed strings needing to be encoded before hashing with md5

This is only necessary under python 3.

Fixed usage of string module to be compatible with python3

Most of the time this consists of calling the function directly from the string.

Replaced usage of subprocess.mswindows

It's an internal API member so we shouldn't be using it.
Also it was removed in Python 3.5.

Fixed a comment causing a syntax error in python3

Python3 was trying to interpret the '\Uxxxx' parts of the comment and failing.

Made sys.platform checks for linux compatible with python3


Made usage of division consistent between python 2 and 3


Replaced usage of apply with argument unpacking syntax

This improves compatibility with python3.

Made all inheritances of 'object' explicit

Also removed a backup declaration of object for python <2.2

Replaced cmp usage with python3 compatible comparisons


Replaced dict.has_key(item) usage with 'item in dict' syntax


Made type comparisons compatible with python3


Made map, reduce, and filter usage compatible with python 3


Made xrange usage compatible with python 3

This is done by either replacing xrange usage with enumerate where appropriate,
or simply by replacing the xrange with range.

Made basestring usage compatible with python 3


Made usage of octals compatible with python 3


Made imports compatible with python 3


Replaced old python2 exception syntax with 'except as' syntax


Made dictionary iterations compatible with python 2 and 3

This mainly consists of replacing, for example, dict.iteritems with dict.items.

In the cases where a list of keys is being generated to get the first item from
the list, this usage has been replaced by converting the dictionary to an
iterator and getting the next item from the iterator.
This also means that we aren't generating a whole list just to get a single
item.

Print statements now come from __future__


Replaced deprecated compiler module usage with the ast module


Replaced deprecated os.path.walk usage with os.walk

Also removed Collector because os.walk can visit directories bottom-up,
so Collector is not needed anymore.

Replaced custom OrderedDict with collections.OrderedDict
==========

to

==========
Well this ended up being bigger than I anticipated!
This changeset makes gyp compatible with python 3 as well as python 2.
This is done without introducing and new module dependencies (such as six).

As it's such a large changeset it may be easier to view it commit by commit,
which you can do here:
https://github.com/AWhetter/gyp/compare/python3Compat...python3Compatibility
If this needs splitting into more changesets to make it more manageable then let
me know!

I've tried to make the changes in here mainly syntax changes. I've put more
"functional" changes into other changesets:
Issue 1406163007
Issue 1429933006
Issue 1429683005
==========

[AWhetter, 2015-11-16 21:41:24]

Description was changed from

==========
Well this ended up being bigger than I anticipated!
This changeset makes gyp compatible with python 3 as well as python 2.
This is done without introducing and new module dependencies (such as six).

As it's such a large changeset it may be easier to view it commit by commit,
which you can do here:
https://github.com/AWhetter/gyp/compare/python3Compat...python3Compatibility
If this needs splitting into more changesets to make it more manageable then let
me know!

I've tried to make the changes in here mainly syntax changes. I've put more
"functional" changes into other changesets:
Issue 1406163007
Issue 1429933006
Issue 1429683005
==========

to

==========
Well this ended up being bigger than I anticipated!
This changeset makes gyp compatible with python 3 as well as python 2.
This is done without introducing and new module dependencies (such as six).

As it's such a large changeset it may be easier to view it commit by commit,
which you can do here:
https://github.com/AWhetter/gyp/compare/python3Compat...python3Compatibility
If this needs splitting into more changesets to make it more manageable then let
me know!

I've tried to make the changes in here mainly syntax changes. I've put more
"functional" changes into other changesets:
Issue #1406163007
Issue 1429933006
Issue 1429683005
==========

[AWhetter, 2015-11-16 21:55:59]

Description was changed from

==========
Well this ended up being bigger than I anticipated!
This changeset makes gyp compatible with python 3 as well as python 2.
This is done without introducing and new module dependencies (such as six).

As it's such a large changeset it may be easier to view it commit by commit,
which you can do here:
https://github.com/AWhetter/gyp/compare/python3Compat...python3Compatibility
If this needs splitting into more changesets to make it more manageable then let
me know!

I've tried to make the changes in here mainly syntax changes. I've put more
"functional" changes into other changesets:
Issue #1406163007
Issue 1429933006
Issue 1429683005
==========

to

==========
Well this ended up being bigger than I anticipated!
As well as being compatible with python 2, this changeset makes gyp (almost)
compatible with python 3 as well.
This is done without introducing any new module dependencies.

As it's such a large changeset it may be easier to view it commit by commit,
which you can do here:
https://github.com/AWhetter/gyp/compare/python3Compat...python3Compatibility
If this needs splitting into more changesets to make it more manageable then let
me know! I realise that this is a big change to land in one go.

I've tried to make the changes in here mainly syntax changes. I've put more
"functional" changes into other changesets:
Replaced custom OrderedDict with collections.OrderedDict
(https://codereview.chromium.org/1429683005)
Replaced deprecated os.path.walk usage with os.walk
(https://codereview.chromium.org/1429933006)
Replaced deprecated compiler module usage with the ast module
(https://codereview.chromium.org/1406163007)

There are some tests that still fail under python 3 but not under python 2.
The test/variables/commands tests fail because the order of iteration of the
parsed dictionary is not guaranteed. In python 2 this never seemed to be an
issue but in python 3 the iteration order does changes intermittently. This in
turn causes the print order of debug messages to change, and therefore not
always match against the expected output. The way this test is checked needs to
be reworked to pass under python 3.
The other tests that fail under python 3 are the MSVS settings tests. This is a
genuine test fail that is again caused by the order of dictionary iteration. On
line 466 of pylib/gyp/MSVSSettings.py we iterate over the msvs_tool_settings
items. The 'DefaultCharIsUnsigned' value requires 'AdditionalOptions' to have
been converted first so that it resides in the msbuild_settings dictionary under
the appropriate tool. A quick and dirty fix is to make
msvs_tool_settings.items() sorted on this line.
I wasn't sure how best to go about resolving these issues so they're up for
discussion.

Fixes issue #36 (https://bugs.chromium.org/p/gyp/issues/detail?id=36)
==========

[AWhetter, 2015-11-16 21:58:51]

Description was changed from

==========
Well this ended up being bigger than I anticipated!
As well as being compatible with python 2, this changeset makes gyp (almost)
compatible with python 3 as well.
This is done without introducing any new module dependencies.

As it's such a large changeset it may be easier to view it commit by commit,
which you can do here:
https://github.com/AWhetter/gyp/compare/python3Compat...python3Compatibility
If this needs splitting into more changesets to make it more manageable then let
me know! I realise that this is a big change to land in one go.

I've tried to make the changes in here mainly syntax changes. I've put more
"functional" changes into other changesets:
Replaced custom OrderedDict with collections.OrderedDict
(https://codereview.chromium.org/1429683005)
Replaced deprecated os.path.walk usage with os.walk
(https://codereview.chromium.org/1429933006)
Replaced deprecated compiler module usage with the ast module
(https://codereview.chromium.org/1406163007)

There are some tests that still fail under python 3 but not under python 2.
The test/variables/commands tests fail because the order of iteration of the
parsed dictionary is not guaranteed. In python 2 this never seemed to be an
issue but in python 3 the iteration order does changes intermittently. This in
turn causes the print order of debug messages to change, and therefore not
always match against the expected output. The way this test is checked needs to
be reworked to pass under python 3.
The other tests that fail under python 3 are the MSVS settings tests. This is a
genuine test fail that is again caused by the order of dictionary iteration. On
line 466 of pylib/gyp/MSVSSettings.py we iterate over the msvs_tool_settings
items. The 'DefaultCharIsUnsigned' value requires 'AdditionalOptions' to have
been converted first so that it resides in the msbuild_settings dictionary under
the appropriate tool. A quick and dirty fix is to make
msvs_tool_settings.items() sorted on this line.
I wasn't sure how best to go about resolving these issues so they're up for
discussion.

Fixes issue #36 (https://bugs.chromium.org/p/gyp/issues/detail?id=36)
==========

to

==========
Well this ended up being bigger than I anticipated!
As well as being compatible with python 2, this changeset makes gyp (almost)
compatible with python 3 as well.
This is done without introducing any new module dependencies.

As it's such a large changeset it may be easier to view it commit by commit,
which you can do here:
https://github.com/AWhetter/gyp/compare/python3Compat...python3Compatibility
If this needs splitting into more changesets to make it more manageable then let
me know! I realise that this is a big change to land in one go.
Half of the line changes come just from updating print statements to include
brackets!

I've tried to make the changes in here mainly syntax changes. I've put more
"functional" changes into other changesets:
Replaced custom OrderedDict with collections.OrderedDict
(https://codereview.chromium.org/1429683005)
Replaced deprecated os.path.walk usage with os.walk
(https://codereview.chromium.org/1429933006)
Replaced deprecated compiler module usage with the ast module
(https://codereview.chromium.org/1406163007)

There are some tests that still fail under python 3 but not under python 2.
The test/variables/commands tests fail because the order of iteration of the
parsed dictionary is not guaranteed. In python 2 this never seemed to be an
issue but in python 3 the iteration order does changes intermittently. This in
turn causes the print order of debug messages to change, and therefore not
always match against the expected output. The way this test is checked needs to
be reworked to pass under python 3.
The other tests that fail under python 3 are the MSVS settings tests. This is a
genuine test fail that is again caused by the order of dictionary iteration. On
line 466 of pylib/gyp/MSVSSettings.py we iterate over the msvs_tool_settings
items. The 'DefaultCharIsUnsigned' value requires 'AdditionalOptions' to have
been converted first so that it resides in the msbuild_settings dictionary under
the appropriate tool. A quick and dirty fix is to make
msvs_tool_settings.items() sorted on this line.
I wasn't sure how best to go about resolving these issues so they're up for
discussion.

Fixes issue #36 (https://bugs.chromium.org/p/gyp/issues/detail?id=36)
==========

[AWhetter, 2015-11-16 22:03:11]

Well this ended up being bigger than I anticipated!

[AWhetter, 2015-11-16 22:10:09]

Description was changed from

==========
Well this ended up being bigger than I anticipated!
As well as being compatible with python 2, this changeset makes gyp (almost)
compatible with python 3 as well.
This is done without introducing any new module dependencies.

As it's such a large changeset it may be easier to view it commit by commit,
which you can do here:
https://github.com/AWhetter/gyp/compare/python3Compat...python3Compatibility
If this needs splitting into more changesets to make it more manageable then let
me know! I realise that this is a big change to land in one go.
Half of the line changes come just from updating print statements to include
brackets!

I've tried to make the changes in here mainly syntax changes. I've put more
"functional" changes into other changesets:
Replaced custom OrderedDict with collections.OrderedDict
(https://codereview.chromium.org/1429683005)
Replaced deprecated os.path.walk usage with os.walk
(https://codereview.chromium.org/1429933006)
Replaced deprecated compiler module usage with the ast module
(https://codereview.chromium.org/1406163007)

There are some tests that still fail under python 3 but not under python 2.
The test/variables/commands tests fail because the order of iteration of the
parsed dictionary is not guaranteed. In python 2 this never seemed to be an
issue but in python 3 the iteration order does changes intermittently. This in
turn causes the print order of debug messages to change, and therefore not
always match against the expected output. The way this test is checked needs to
be reworked to pass under python 3.
The other tests that fail under python 3 are the MSVS settings tests. This is a
genuine test fail that is again caused by the order of dictionary iteration. On
line 466 of pylib/gyp/MSVSSettings.py we iterate over the msvs_tool_settings
items. The 'DefaultCharIsUnsigned' value requires 'AdditionalOptions' to have
been converted first so that it resides in the msbuild_settings dictionary under
the appropriate tool. A quick and dirty fix is to make
msvs_tool_settings.items() sorted on this line.
I wasn't sure how best to go about resolving these issues so they're up for
discussion.

Fixes issue #36 (https://bugs.chromium.org/p/gyp/issues/detail?id=36)
==========

to

==========
Well this ended up being bigger than I anticipated!
As well as being compatible with python 2, this changeset makes gyp (almost)
compatible with python 3 as well.
This is done without introducing any new module dependencies.

As it's such a large changeset it may be easier to view it commit by commit,
which you can do here:
https://github.com/AWhetter/gyp/compare/python3Compat...python3Compatibility
If this needs splitting into more changesets to make it more manageable then let
me know! I realise that this is a big change to land in one go.
Half of the line changes come just from updating print statements to include
brackets!
It's maybe worth looking at the commits anyway because I've tried to make the
commit messages describe my thinking behind the change.

I've tried to make the changes in here mainly syntax changes. I've put more
"functional" changes into other changesets:
Replaced custom OrderedDict with collections.OrderedDict
(https://codereview.chromium.org/1429683005)
Replaced deprecated os.path.walk usage with os.walk
(https://codereview.chromium.org/1429933006)
Replaced deprecated compiler module usage with the ast module
(https://codereview.chromium.org/1406163007)

There are some tests that still fail under python 3 but not under python 2.
The test/variables/commands tests fail because the order of iteration of the
parsed dictionary is not guaranteed. In python 2 this never seemed to be an
issue but in python 3 the iteration order does changes intermittently. This in
turn causes the print order of debug messages to change, and therefore not
always match against the expected output. The way this test is checked needs to
be reworked to pass under python 3.
The other tests that fail under python 3 are the MSVS settings tests. This is a
genuine test fail that is again caused by the order of dictionary iteration. On
line 466 of pylib/gyp/MSVSSettings.py we iterate over the msvs_tool_settings
items. The 'DefaultCharIsUnsigned' value requires 'AdditionalOptions' to have
been converted first so that it resides in the msbuild_settings dictionary under
the appropriate tool. A quick and dirty fix is to make
msvs_tool_settings.items() sorted on this line.
I wasn't sure how best to go about resolving these issues so they're up for
discussion.

Fixes issue #36 (https://bugs.chromium.org/p/gyp/issues/detail?id=36)
==========

[Nico, 2016-07-29 22:22:03]

thakis@chromium.org changed reviewers:
+ thakis@chromium.org

[Nico, 2016-07-29 22:22:06]

Thanks for the patch, this looks really good. A few questions below, I commented
on each instance only once. (Ignore the first comment.)

My mindset while looking through this was mostly "will things still work with
python 2.7 after this?"

To get a review, you normally put someone in the "reviewers" line; if you just
cc gyp-developer nobody will look at patches I'm afraid.

https://codereview.chromium.org/1454433002/diff/20001/pylib/gyp/MSVSNew.py
File pylib/gyp/MSVSNew.py (right):

https://codereview.chromium.org/1454433002/diff/20001/pylib/gyp/MSVSNew.py#ne...
pylib/gyp/MSVSNew.py:64: def __lt__(self, other):
This looks like an unfortunate change :-/

https://codereview.chromium.org/1454433002/diff/20001/pylib/gyp/MSVSSettings_...
File pylib/gyp/MSVSSettings_test.py (right):

https://codereview.chromium.org/1454433002/diff/20001/pylib/gyp/MSVSSettings_...
pylib/gyp/MSVSSettings_test.py:12: from io import StringIO
Isn't the recommended pattern "try py3 way, and if that fails fall back to py2"?

https://codereview.chromium.org/1454433002/diff/20001/pylib/gyp/common.py
File pylib/gyp/common.py (right):

https://codereview.chromium.org/1454433002/diff/20001/pylib/gyp/common.py#new...
pylib/gyp/common.py:348: self.tmp_file = os.fdopen(tmp_fd, 'w')
doesn't this change behavior on windows?

https://codereview.chromium.org/1454433002/diff/20001/pylib/gyp/input.py
File pylib/gyp/input.py (right):

https://codereview.chromium.org/1454433002/diff/20001/pylib/gyp/input.py#newc...
pylib/gyp/input.py:1209: for key, value in the_dict.items():
Did you measure if this change makes gyp slower when running with 2.7?

https://codereview.chromium.org/1454433002/diff/20001/test/actions-multiple/s...
File test/actions-multiple/src/actions.gyp (right):

https://codereview.chromium.org/1454433002/diff/20001/test/actions-multiple/s...
test/actions-multiple/src/actions.gyp:29: 'inputs': [
why this change?

https://codereview.chromium.org/1454433002/diff/20001/test/errors/gyptest-err...
File test/errors/gyptest-errors.py (right):

https://codereview.chromium.org/1454433002/diff/20001/test/errors/gyptest-err...
test/errors/gyptest-errors.py:44: "target2.*target0.*target1.*target2).*")
why this change?

https://codereview.chromium.org/1454433002/diff/20001/test/lib/TestGyp.py
File test/lib/TestGyp.py (right):

https://codereview.chromium.org/1454433002/diff/20001/test/lib/TestGyp.py#new...
test/lib/TestGyp.py:891: C:/USERS/yourname/AppData/Local. It will cause needless
work.  The ouptut
out of interest, is this needed? with \ it's more convenient to paste into
cmd.exe i suppose

https://codereview.chromium.org/1454433002/diff/20001/test/variables/commands...
File test/variables/commands/commands.gyp (right):

https://codereview.chromium.org/1454433002/diff/20001/test/variables/commands...
test/variables/commands/commands.gyp:11: 'pi': 'import math; print(math.pi)',
hmm doesn't this need an `from __future__ import print_function` too when
running gyp under python 2.7?

[AWhetter, 2016-11-05 23:59:50]

I've rebased with master again and pushed these changes to github. Let me know
if you think it needs more work!

https://codereview.chromium.org/1454433002/diff/20001/pylib/gyp/MSVSNew.py
File pylib/gyp/MSVSNew.py (right):

https://codereview.chromium.org/1454433002/diff/20001/pylib/gyp/MSVSNew.py#ne...
pylib/gyp/MSVSNew.py:64: def __lt__(self, other):
On 2016/07/29 at 22:22:05, Nico (mostly afk Wed-Fri) wrote:
> This looks like an unfortunate change :-/

It is really. I could make it slighty easier on the eyes and do something like
add a _cmp method that takes an operator as an argument, and __lt__, __gt__, etc
call that with an operator.

https://codereview.chromium.org/1454433002/diff/20001/pylib/gyp/MSVSSettings_...
File pylib/gyp/MSVSSettings_test.py (right):

https://codereview.chromium.org/1454433002/diff/20001/pylib/gyp/MSVSSettings_...
pylib/gyp/MSVSSettings_test.py:12: from io import StringIO
On 2016/07/29 at 22:22:05, Nico (mostly afk Wed-Fri) wrote:
> Isn't the recommended pattern "try py3 way, and if that fails fall back to
py2"?

You're right. I've fixed this on my github.

https://codereview.chromium.org/1454433002/diff/20001/pylib/gyp/common.py
File pylib/gyp/common.py (right):

https://codereview.chromium.org/1454433002/diff/20001/pylib/gyp/common.py#new...
pylib/gyp/common.py:348: self.tmp_file = os.fdopen(tmp_fd, 'w')
On 2016/07/29 at 22:22:06, Nico (mostly afk Wed-Fri) wrote:
> doesn't this change behavior on windows?

Yes but we only ever write text so I think we should be doing this anyway?

https://codereview.chromium.org/1454433002/diff/20001/pylib/gyp/input.py
File pylib/gyp/input.py (right):

https://codereview.chromium.org/1454433002/diff/20001/pylib/gyp/input.py#newc...
pylib/gyp/input.py:1209: for key, value in the_dict.items():
On 2016/07/29 at 22:22:06, Nico (mostly afk Wed-Fri) wrote:
> Did you measure if this change makes gyp slower when running with 2.7?

Just with a few runs of all of the tests, I got the same run time of 2m50s for
both with and without the changes. My guess is the dictionaries are small enough
for the switch from list to generator to not make a big difference.

https://codereview.chromium.org/1454433002/diff/20001/test/actions-multiple/s...
File test/actions-multiple/src/actions.gyp (right):

https://codereview.chromium.org/1454433002/diff/20001/test/actions-multiple/s...
test/actions-multiple/src/actions.gyp:29: 'inputs': [
On 2016/07/29 at 22:22:06, Nico (ooo until Aug 1) wrote:
> why this change?

I renamed copy.py to copyfile.py because
> Down the import dependency chain of copy.py the 'tarfile' module was trying to
import the 'copy' standard library module but was finding the copy.py file first
because it was in the current directory.

I get the following error with Python 3 if I don't rename the file:
Traceback (most recent call last):
  File "copy.py", line 6, in <module>
    import shutil
  File "/usr/lib/python3.5/shutil.py", line 13, in <module>
    import tarfile
  File "/usr/lib/python3.5/tarfile.py", line 49, in <module>
    import copy
  File
"/home/ashley/workspace/gyp/out/testworkarea/testgyp.giniq4wt/relocate/src/copy.py",
line 9, in <module>
    shutil.copyfile(sys.argv[1], sys.argv[2])
AttributeError: module 'shutil' has no attribute 'copyfile'

https://codereview.chromium.org/1454433002/diff/20001/test/errors/gyptest-err...
File test/errors/gyptest-errors.py (right):

https://codereview.chromium.org/1454433002/diff/20001/test/errors/gyptest-err...
test/errors/gyptest-errors.py:44: "target2.*target0.*target1.*target2).*")
On 2016/07/29 at 22:22:06, Nico (ooo until Aug 1) wrote:
> why this change?

To quote the commit message:
> The order in which the nodes of a cycle are printed is unpredictable. I
believe this is because the order of iteration over the items in the dependency
dictionary is unpredictable. Therefore the error tests now match against all the
possible orderings of dependencies in a detected cycle.

https://codereview.chromium.org/1454433002/diff/20001/test/lib/TestGyp.py
File test/lib/TestGyp.py (right):

https://codereview.chromium.org/1454433002/diff/20001/test/lib/TestGyp.py#new...
test/lib/TestGyp.py:891: C:/USERS/yourname/AppData/Local. It will cause needless
work.  The ouptut
On 2016/07/29 at 22:22:06, Nico (mostly afk Wed-Fri) wrote:
> out of interest, is this needed? with \ it's more convenient to paste into
cmd.exe i suppose

I've worked around it by making the docstring a raw string instead. So they're
back to being backslashes again.

https://codereview.chromium.org/1454433002/diff/20001/test/variables/commands...
File test/variables/commands/commands.gyp (right):

https://codereview.chromium.org/1454433002/diff/20001/test/variables/commands...
test/variables/commands/commands.gyp:11: 'pi': 'import math; print(math.pi)',
On 2016/07/29 at 22:22:06, Nico (mostly afk Wed-Fri) wrote:
> hmm doesn't this need an `from __future__ import print_function` too when
running gyp under python 2.7?

Nope. Using brackets is optional in Python 2.7.

>>> print 'test'
test
>>> print('test')
test
>>> print 'Hello', 'world'
Hello world
>>> print('Hello', 'world')
('Hello', 'world')