ninja windows: fix mapping of optimization flags

ninja windows: fix mapping of optimization flags

"FavorSizeOrSpeed" should map to Os/Ot. These were backwards.

"Optimization" can also be set to Disabled/Size/Speed/Max which maps to
Od/O1/O2/Ox. Seems a bit silly, but that's how VS exposes the compiler
settings. Previously, Size/Speed here were incorrectly duplicating the Os/Ot
setting rather than O1/O2.

So, before we were basically doing some random set of optimizing flags.

(I have no idea what the compiler actually attempts to do when you set O1+Ot
or O2+Os. I guess it must be macro vs. micro level, but it's not very clear
from the names. Chromium sets them consistently.)

R=thakis@chromium.org
BUG=126315
Committed: https://code.google.com/p/gyp/source/detail?r=1363

[Nico, 2012-05-06 20:16:47]

LGTM, thanks for looking into this :-)

(ps: I note that this happened in the one test file that doesn't compare results
with cl.exe)

https://chromiumcodereview.appspot.com/10381032/diff/3001/test/win/compiler-f...
File test/win/compiler-flags/optimizations.gyp (right):

https://chromiumcodereview.appspot.com/10381032/diff/3001/test/win/compiler-f...
test/win/compiler-flags/optimizations.gyp:125: },
nit: I read http://msdn.microsoft.com/en-us/library/aa652267(v=vs.71).aspx to
review this; maybe there should be a test for '0' too. Not very important.

https://chromiumcodereview.appspot.com/10381032/diff/3001/test/win/gyptest-cl...
File test/win/gyptest-cl-optimizations.py (right):

https://chromiumcodereview.appspot.com/10381032/diff/3001/test/win/gyptest-cl...
test/win/gyptest-cl-optimizations.py:40: test.must_not_contain(ninja_file,
'/Od')
nit: Od, O1, O2, Ox (to match order above)

[scottmg, 2012-05-07 00:58:04]

On 2012/05/06 20:16:47, Nico wrote:
> LGTM, thanks for looking into this :-)
> 
> (ps: I note that this happened in the one test file that doesn't compare
results
> with cl.exe)

Yeah :/ I looked at it a little, but with all the various flags it gets very
complex. Maybe if we only supported one Windows compiler version instead of 2-3
versions it'd be reasonable.

> 
>
https://chromiumcodereview.appspot.com/10381032/diff/3001/test/win/compiler-f...
> File test/win/compiler-flags/optimizations.gyp (right):
> 
>
https://chromiumcodereview.appspot.com/10381032/diff/3001/test/win/compiler-f...
> test/win/compiler-flags/optimizations.gyp:125: },
> nit: I read http://msdn.microsoft.com/en-us/library/aa652267%28v=vs.71%29.aspx
to
> review this; maybe there should be a test for '0' too. Not very important.

Done.

> 
>
https://chromiumcodereview.appspot.com/10381032/diff/3001/test/win/gyptest-cl...
> File test/win/gyptest-cl-optimizations.py (right):
> 
>
https://chromiumcodereview.appspot.com/10381032/diff/3001/test/win/gyptest-cl...
> test/win/gyptest-cl-optimizations.py:40: test.must_not_contain(ninja_file,
> '/Od')
> nit: Od, O1, O2, Ox (to match order above)

Done.