xcode emulation for make: Adjust .a and .dylib entries in 'libraries'.

xcode emulation for make: Adjust .a and .dylib entries in 'libraries'.

Transform, e.g., 'libfoo.a' to '-lfoo', so that the LIBRARY_SEARCH_PATHS are
actually searched for libfoo.a. (Warning: This could have the side effect that
libfoo.dylib is preferred over libfoo.a.)

TEST=test/mac/gyptest-libraries.py
Committed: https://code.google.com/p/gyp/source/detail?r=1173

[Nico, 2012-01-19 19:59:08]

needs test

[Nico, 2012-01-19 20:06:11]

(The reason for that is mostly to make sure that the behavior does really match
xcode. I think xcode adds libraries as libraries, and not as -l if I remember
correctly.)

[viettrungluu, 2012-01-19 21:43:43]

On 2012/01/19 20:06:11, Nico wrote:
> (The reason for that is mostly to make sure that the behavior does really
match
> xcode. I think xcode adds libraries as libraries, and not as -l if I remember
> correctly.)

Test added, PTAL. You remember incorrectly. Or incompletely. Or whatever. It
really does add libfoo.{a,dylib} as -lfoo. I'm guessing it doesn't touch, e.g.,
<path>/libfoo.a though.

[Nico, 2012-01-20 17:44:06]

http://codereview.chromium.org/9254033/diff/6001/pylib/gyp/xcode_emulation.py
File pylib/gyp/xcode_emulation.py (right):

http://codereview.chromium.org/9254033/diff/6001/pylib/gyp/xcode_emulation.py...
pylib/gyp/xcode_emulation.py:386: ldflags.append('-L' +
gyp_to_build_path(library_path))
Can you add a test for this, too? (That's not a question.)

http://codereview.chromium.org/9254033/diff/6001/pylib/gyp/xcode_emulation.py...
pylib/gyp/xcode_emulation.py:552: libraries = [ self._AdjustLibrary(library) for
library in libraries]
I like this, but linux doesn't do it. Should we be doing this on linux too?
(cc'd gyp-developer list, any opinion?)

("this" being the translation of libfoo.a to -lfoo)

[viettrungluu, 2012-01-21 00:15:19]

http://codereview.chromium.org/9254033/diff/6001/pylib/gyp/xcode_emulation.py
File pylib/gyp/xcode_emulation.py (right):

http://codereview.chromium.org/9254033/diff/6001/pylib/gyp/xcode_emulation.py...
pylib/gyp/xcode_emulation.py:386: ldflags.append('-L' +
gyp_to_build_path(library_path))
On 2012/01/20 17:44:07, Nico wrote:
> Can you add a test for this, too? (That's not a question.)

I added a test for LIBRARY_SEARCH_PATHS, but it doesn't actually manage to
verify that the path is properly transformed.

Where do you suggest that I copy a library to in order to check that
gyp_to_build_path() is properly called?

postbuilds/actions support is pretty broken, btw, with lots of variables that
xcodebuild exports not being exported. This isn't making my life any easier.

http://codereview.chromium.org/9254033/diff/6001/pylib/gyp/xcode_emulation.py...
pylib/gyp/xcode_emulation.py:552: libraries = [ self._AdjustLibrary(library) for
library in libraries]
On 2012/01/20 17:44:07, Nico wrote:
> I like this, but linux doesn't do it. Should we be doing this on linux too?

No .gyp file for Linux currently relies on this. Though it'd be more consistent
with the other generators/platforms.

> (cc'd gyp-developer list, any opinion?)
> 
> ("this" being the translation of libfoo.a to -lfoo)

[Nico, 2012-01-21 00:18:15]

http://codereview.chromium.org/9254033/diff/6001/pylib/gyp/xcode_emulation.py
File pylib/gyp/xcode_emulation.py (right):

http://codereview.chromium.org/9254033/diff/6001/pylib/gyp/xcode_emulation.py...
pylib/gyp/xcode_emulation.py:386: ldflags.append('-L' +
gyp_to_build_path(library_path))
On 2012/01/21 00:15:20, viettrungluu wrote:
> On 2012/01/20 17:44:07, Nico wrote:
> > Can you add a test for this, too? (That's not a question.)
> 
> I added a test for LIBRARY_SEARCH_PATHS, but it doesn't actually manage to
> verify that the path is properly transformed.
> 
> Where do you suggest that I copy a library to in order to check that
> gyp_to_build_path() is properly called?

Add a 'static_library' target, then from the test build this, and move it to a
subdirectory of product_dir (from the test code) before building the target that
uses LIBRARY_SEARCH_PATHS.

> postbuilds/actions support is pretty broken, btw, with lots of variables that
> xcodebuild exports not being exported. This isn't making my life any easier.

See the "as needed" comment at the top of GetXcodeEnv() :-)

[viettrungluu, 2012-01-21 00:40:08]

Too much trouble. Cutting my losses.

[viettrungluu, 2012-01-23 22:00:58]

Apparently, this yak was nearly shorn anyway. And the other yak was dead. PTAL.

[Nico, 2012-01-23 22:05:47]

As said above, I feel like this should work with all generators on all
platforms. But doing that is fine in a follow-up (not necessarily by you), and
this matches xcode, so LGTM.

https://chromiumcodereview.appspot.com/9254033/diff/14001/test/mac/gyptest-li...
File test/mac/gyptest-libraries.py (right):

https://chromiumcodereview.appspot.com/9254033/diff/14001/test/mac/gyptest-li...
test/mac/gyptest-libraries.py:16: # TODO: Add ninja once it supports postbuilds.
nit: all tests without 'ninja' have an implicit TODO on me to enable them, so
you can remove this if you want.

https://chromiumcodereview.appspot.com/9254033/diff/14001/test/mac/libraries/...
File test/mac/libraries/subdir/hello.cc (right):

https://chromiumcodereview.appspot.com/9254033/diff/14001/test/mac/libraries/...
test/mac/libraries/subdir/hello.cc:8: std::cout << "Hello, world!" << std::endl;
y u no use stdio :-/

[Nico, 2012-01-23 22:05:58]

…and thanks for not giving up :-)