ninja/mac: Don't write an empty binary into sourceless bundles.

ninja/mac: Don't write an empty binary into sourceless bundles.

I fixed this issue once in https://codereview.chromium.org/9121011/ but added
the test in the wrong way so that it's not run automatically, and
https://codereview.chromium.org/9107028 broke that again.

Rename the test from .gyp to .py to make sure the bots run it, so that it
doesn't break again.

BUG=chromium:280718
R=mark@chromium.org

Committed: https://code.google.com/p/gyp/source/detail?r=1709

[Nico, 2013-09-03 19:47:20]

I'd send this to Scott, but he's out until tomorrow. It's mostly a no-op change.

In WriteTarget(), if there are no link_deps (i.e. no .o files), handle the
target like a type none target (which just produces a stamp file for
dependencies, instead of calling ld). This is the more interesting (and simpler)
change.

In WriteMacBundleOutput(), this now writes output.stamp file instead of touching
the bundle's name if the bundle has neither bundle resources nor a binary. This
isn't terribly important, but matches xcode's behavior so it's nice to have I
suppose (and the existing test tests for this)

[Mark Mentovai, 2013-09-03 20:00:48]

LGTM

[Nico, 2013-09-03 20:03:14]

Committed patchset #2 manually as r1709 (presubmit successful).

[justincohen, 2013-09-06 14:06:05]

https://codereview.chromium.org/23916002/diff/7001/pylib/gyp/generator/ninja.py
File pylib/gyp/generator/ninja.py (right):

https://codereview.chromium.org/23916002/diff/7001/pylib/gyp/generator/ninja....
pylib/gyp/generator/ninja.py:482: is_empty_bundle = not link_deps and not
mac_bundle_depends
This breaks some bundles we depend on that are only mac_bundle_resources.  In
this case, mac_bundle_depends is not nil, but since line 478 evaluates to False,
this will never run.  Is there a reason this is nested inside #478?  Why not
just check for not link_deps and not mac_bundle_depends on line 477?

This CL breaks ninja for iOS builds.  Perhaps one of the bundles we depend on is
doing something wrong?

[Nico, 2013-09-06 14:10:16]

https://codereview.chromium.org/23916002/diff/7001/pylib/gyp/generator/ninja.py
File pylib/gyp/generator/ninja.py (right):

https://codereview.chromium.org/23916002/diff/7001/pylib/gyp/generator/ninja....
pylib/gyp/generator/ninja.py:482: is_empty_bundle = not link_deps and not
mac_bundle_depends
On 2013/09/06 14:06:05, justincohen wrote:
> This breaks some bundles we depend on that are only mac_bundle_resources.  In
> this case, mac_bundle_depends is not nil, but since line 478 evaluates to
False,
> this will never run.  Is there a reason this is nested inside #478?  Why not
> just check for not link_deps and not mac_bundle_depends on line 477?

Sounds reasonable.

> This CL breaks ninja for iOS builds.  Perhaps one of the bundles we depend on
is
> doing something wrong?

Not sure, this CL explicitly is about supporting sourceless bundles (i.e. just
mac_bundle_resources) better, see the tests. I'm guessing your sourceless
bundles look slightly different? I suggest adding a demo of your problem to
test/mac/sourceless-bundle/test.gyp / test/mac/gyptest-sourceless-bundle.py and
then sending a CL with that test and your proposed fix.