Change cxxflags test to build in same env as gyp (+ test make instead of ninja)

Change cxxflags test to build in same env as gyp (+ test make instead of ninja)

Closed:
handled by using a define instead:
Make test/cxxflags use a user define instead of __OPTIMIZE__
https://codereview.chromium.org/235033003/

Fixes test/build breakage in GYP r1893
https://code.google.com/p/gyp/source/detail?r=1893
http://build.chromium.org/p/tryserver.nacl/builders/gyp-win32/builds/1710

...due to different behavior on ninja-Linux and ninja-Windows (apparently).

Changes:
1. Switch back to testing make, instead of ninja.
(Test originally only tested make.)
2. Changes test to always use same build env as gyp, not different one,
as otherwise make fails to use gyp time settings.
(Different build env from gyp has build-system and platform dependent results:
works on ninja-Linux, for instance, but not make.)

Follow-up to:
Use context manager to manage OS environment in tests
https://codereview.chromium.org/228323002/

R=thakis
TEST=PYTHONPATH=$CHROMIUM_DIR/tools/gyp/gyptest.py -f make test/cxxflags/gyptest-cxxflags.py

[Nils Barth (inactive), 2014-04-11 02:42:34]

Heyas,
Here's a quick CL that should fix the Windows build/test sadness.
If this is the fix you want, and it works ok on Windows,
please feel free to go ahead and commit it!

Thanks.

[scottmg, 2014-04-11 03:01:53]

I still get this on Windows:

d:\src\cr3\src\tools\gyp>python gyptest.py -f ninja
test\cxxflags\gyptest-cxxflags.py
PYTHONPATH=d:\src\cr3\src\tools\gyp\test\lib
TESTGYP_FORMAT=ninja
d:\src\depot_tools\python276_bin\python.exe test\cxxflags\gyptest-cxxflags.py
STDOUT =========================================================================
1c1
< Using an optimization flag
---
> Using no optimization flag
FAILED test at line 429 of d:\src\cr3\src\tools\gyp\test\lib\TestCommon.py
(_complete)
        from line 532 of d:\src\cr3\src\tools\gyp\test\lib\TestCommon.py (run)
        from line 293 of d:\src\cr3\src\tools\gyp\test\lib\TestGyp.py (run)
        from line 873 of d:\src\cr3\src\tools\gyp\test\lib\TestGyp.py
(run_built_executable)
        from line 43 of test\cxxflags\gyptest-cxxflags.py

Failed the following test:
        test\cxxflags\gyptest-cxxflags.py


If you're not sure what's going on/don't have win, I can look at it tomorrow.

[Nils Barth (inactive), 2014-04-11 04:31:44]

On 2014/04/11 03:01:53, scottmg wrote:
> I still get this on Windows:
> 
> d:\src\cr3\src\tools\gyp>python gyptest.py -f ninja
> test\cxxflags\gyptest-cxxflags.py

Ok, looks like ninja was disabled for a reason.
Originally the test only tested make, not ninja.

I switched it to ninja b/c make failed once I corrected the test
(so build occurred in different env from gyp),
but looks like the only workable fix is to have build and gyp
in same environment.

> If you're not sure what's going on/don't have win, I can look at it tomorrow.

Could you?
I've posted a fix that should work, if you want the tests to be passing again,
and corresponds to previous test formats (make only).

However, you might want to look into whether:
* make should work if build and gyp env differ
* ninja should work with -O set in CXXFLAGS env on Windows
(-D set in CFLAGS works on Linux/Windows)
(I don't know what's expected.)

Thanks for patience!

[scottmg, 2014-04-11 18:23:31]

On 2014/04/11 04:31:44, Nils Barth wrote:
> On 2014/04/11 03:01:53, scottmg wrote:
> > I still get this on Windows:
> > 
> > d:\src\cr3\src\tools\gyp>python gyptest.py -f ninja
> > test\cxxflags\gyptest-cxxflags.py
> 
> Ok, looks like ninja was disabled for a reason.
> Originally the test only tested make, not ninja.
> 
> I switched it to ninja b/c make failed once I corrected the test
> (so build occurred in different env from gyp),
> but looks like the only workable fix is to have build and gyp
> in same environment.
> 
> > If you're not sure what's going on/don't have win, I can look at it
tomorrow.
> 
> Could you?
> I've posted a fix that should work, if you want the tests to be passing again,
> and corresponds to previous test formats (make only).
> 
> However, you might want to look into whether:
> * make should work if build and gyp env differ
> * ninja should work with -O set in CXXFLAGS env on Windows
> (-D set in CFLAGS works on Linux/Windows)
> (I don't know what's expected.)
> 
> Thanks for patience!

CXXFLAGS appears to work fine, the test just strangely assumes __OPTIMIZE__ will
be set. I'm not aware of any preprocessor macro that will be set on Windows when
optimization is on/off. (It seems odd to me, since that's a codegen thing, not a
preprocessor thing.)

[Nils Barth (inactive), 2014-04-14 01:58:17]

On 2014/04/11 18:23:31, scottmg wrote:
> CXXFLAGS appears to work fine, the test just strangely assumes __OPTIMIZE__
will
> be set. I'm not aware of any preprocessor macro that will be set on Windows
when
> optimization is on/off. (It seems odd to me, since that's a codegen thing, not
a
> preprocessor thing.)

Hmm...so presumably the .cc test file needs fixing?
(Easiest would just be to pass -D as a flag, as in the cflags test.)

Looking at the GCC vs. MSVC predefined macros, I don't think there's any
overlap:
http://gcc.gnu.org/onlinedocs/cpp/Common-Predefined-Macros.html
http://msdn.microsoft.com/en-us/library/b0084kay.aspx
...as the GCC ones all start with __, but only a few of the MSVC ones do,
and they don't overlap.

http://stackoverflow.com/questions/14618249/is-there-a-builtin-macro-defined-...

What should we do?

[Nils Barth (inactive), 2014-04-15 01:17:36]

On 2014/04/14 01:58:17, Nils Barth wrote:
> Hmm...so presumably the .cc test file needs fixing?
> (Easiest would just be to pass -D as a flag, as in the cflags test.)

Yup, done in:
Make test/cxxflags use a user define instead of __OPTIMIZE__
https://codereview.chromium.org/235033003/

Thanks Scott!