win msvs: make ordering match .gyp order

win msvs: make ordering match .gyp order

The order of .obj on the link command line determines the order in the
final binary in the absence of other need to re-order. Make them
consistent across the windows generators by making them match the order
in the .gyp file, and add a test for this behaviour. This was already
ninja's behaviour.

R=thakis@chromium.org
BUG=chromium:326030
Committed: https://code.google.com/p/gyp/source/detail?r=1807

[Nico, 2013-12-06 06:38:24]

Can't both msvs and ninja use the order from the ninja file, wouldn't they match
then too? I think that's the current order in ninja at least.

[scottmg, 2013-12-06 07:05:42]

Now in gyp order instead of sorted.

[Nico, 2013-12-06 16:28:29]

Thanks, I like this much better conceptually. I found a few places that still
look like they use set functions, but maybe I'm misreading the code – if the
tests pass, things should be cool I guess? So lgtm.

From what I can tell, we shouldn't end up with duplicate files in projects with
this change, but if this is a concern, maybe adding a subset (hah!) of
http://code.activestate.com/recipes/576694/ and using that might be an option.

https://codereview.chromium.org/107293004/diff/60001/pylib/gyp/generator/msvs.py
File pylib/gyp/generator/msvs.py (right):

https://codereview.chromium.org/107293004/diff/60001/pylib/gyp/generator/msvs...
pylib/gyp/generator/msvs.py:595: sources.add(filename)
Will this still work?

https://codereview.chromium.org/107293004/diff/60001/pylib/gyp/generator/msvs...
pylib/gyp/generator/msvs.py:805: sources.update(inputs)
Won't this throw now that sources is a set? (I guess it's fine somehow if tests
pass? I don't understand why though.)

https://codereview.chromium.org/107293004/diff/60001/pylib/gyp/generator/msvs...
pylib/gyp/generator/msvs.py:893: def _GenerateMSVSProject(project, options,
version, generator_flags):
Unrelated: This can probably be removed eventually

https://codereview.chromium.org/107293004/diff/60001/pylib/gyp/generator/msvs...
pylib/gyp/generator/msvs.py:1333: def _AddNormalizedSources(sources_list,
sources_array):
sources_list and sources_array, hm? :-)

https://codereview.chromium.org/107293004/diff/60001/pylib/gyp/generator/msvs...
pylib/gyp/generator/msvs.py:1380: Also converts the sets to lists.
remove?

https://codereview.chromium.org/107293004/diff/60001/pylib/gyp/generator/msvs...
pylib/gyp/generator/msvs.py:1387: excluded_sources: A list of sources to be
excluded for this project.
maybe add list_excluded while you're here

[scottmg, 2013-12-06 19:10:22]

Sorry, I should have waited for the tryjobs to finish.

There're cases where e.g. action scripts are re-used which causes them to be
added multiple times so the set is somewhat necessary.

I switched to using an OrderedSet instead per your suggested link which is a
simpler change and also has the benefit of "working".

[Nico, 2013-12-06 20:35:53]

lgtm

https://codereview.chromium.org/107293004/diff/120001/pylib/gyp/generator/msv...
File pylib/gyp/generator/msvs.py (right):

https://codereview.chromium.org/107293004/diff/120001/pylib/gyp/generator/msv...
pylib/gyp/generator/msvs.py:102: def discard(self, key):
Is this needed?

[scottmg, 2013-12-06 20:43:46]

https://codereview.chromium.org/107293004/diff/120001/pylib/gyp/generator/msv...
File pylib/gyp/generator/msvs.py (right):

https://codereview.chromium.org/107293004/diff/120001/pylib/gyp/generator/msv...
pylib/gyp/generator/msvs.py:102: def discard(self, key):
On 2013/12/06 20:35:54, Nico wrote:
> Is this needed?

Yup:
http://docs.python.org/2/library/collections.html#collections-abstract-base-c...
I didn't even know there was such a thing in python. Without them it complains:

    TypeError: Can't instantiate abstract class OrderedSet with abstract methods
__len__, discard

(I could remove the bodies I guess, but that seems like asking for trouble in
the future).

[scottmg, 2013-12-06 21:04:44]

Committed patchset #3 manually as r1807 (presubmit successful).