Xcode has some issues with generated sources/headers and compiling sources. ...

Xcode has some issues with generated sources/headers and compiling sources.  This splits the actions/rules out into support targets to help coax Xcode into rebuilding when it should.  There still seem to be some possible issues, but it does fix some fail cases we were able to create.
Fix the case where test runners were not getting the configuration lists.

TEST=everything still builds, and hopefully things do better at rebuilding when they should.
BUG=none

Committed revision 762.

[TVL, 2009-11-23 19:16:18]

I'm still watching this in my local tree for a few more builds, but I figured
I'd send it out for other eyes at the same time.

[TVL, 2009-11-23 21:10:25]

Ok, cl should be go, also fixed the issue with test runners not getting the
configurations they should.

[sgk, 2009-11-23 21:24:11]

Can the fail cases be captured in test case(s) that we can make part of the
suite?  From the general description, this involves subtle Xcode behavior that
would be easy for future refactorings to break.

[Mark Mentovai, 2009-11-23 21:25:15]

LGTM.

Does this make us pass a test case that we wouldn't pass before, or that we
wouldn't always pass?  Do we need to do something special to reenable that test
case?

http://codereview.chromium.org/437013/diff/2001/2002
File pylib/gyp/generator/xcode.py (left):

http://codereview.chromium.org/437013/diff/2001/2002#oldcode534
pylib/gyp/generator/xcode.py:534: 
Looks like this file uses two blank lines between functions, leave this in.

http://codereview.chromium.org/437013/diff/2001/2002#oldcode680
pylib/gyp/generator/xcode.py:680: # TODO(mark): this assumes too much knowledge
of the internals of
You don't think this TODO is still valid?

http://codereview.chromium.org/437013/diff/2001/2002#oldcode955
pylib/gyp/generator/xcode.py:955: # TODO(mark): this assumes too much knowledge
of the internals of
Same?

http://codereview.chromium.org/437013/diff/2001/2002
File pylib/gyp/generator/xcode.py (right):

http://codereview.chromium.org/437013/diff/2001/2002#newcode644
pylib/gyp/generator/xcode.py:644: # sources" step vs. any source files/headers
generated by actions/rules.
"vs." doesn't mean what you want here.

http://codereview.chromium.org/437013/diff/2001/2002#newcode660
pylib/gyp/generator/xcode.py:660:
gyp.xcodeproj_file.PBXAggregateTarget(support_target_properties,
Same.  And that'll cause the next line to need to be bumped over too.

http://codereview.chromium.org/437013/diff/2001/2002#newcode1072
pylib/gyp/generator/xcode.py:1072: # The support project also gets the
dependencies (incase they are
in case

[TVL, 2009-11-23 21:35:43]

On 2009/11/23 21:24:11, sgk wrote:
> Can the fail cases be captured in test case(s) that we can make part of the
> suite?  From the general description, this involves subtle Xcode behavior that
> would be easy for future refactorings to break.

It should fix http://codereview.chromium.org/300016 which seems like it didn't
land?  Rather then do a new test, can we get this landed?  (although it looks
like it modifies files that get checked out, maybe we should tweak it not to do
that first)

[TVL, 2009-11-23 21:44:50]

comment/indent fixes up.