Allow GYP targets to depend on themselves if they have the 'prune_self_dependencies' set to 1

Allow GYP targets to depend on themselves if they have the 'prune_self_dependencies' set to 1.

See https://codereview.chromium.org/11642018/ for more context.

BUG=chromium:170629

[glider1, 2013-02-25 18:32:23]

Mark, is this the right thing to do?

[glider1, 2013-02-25 18:36:41]

I haven't found any GYP tests related to cycles and will probably need to add
one. Feel free to disregard this patch until then.

[Mark Mentovai, 2013-02-25 18:49:42]

It occurs to me that this will allow at most one target in the entire system to
be made a forced dependency of all other targets. As soon as you add a second
one, they’ll wind up depending on each other cyclically.

That makes this solution imperfect, although it’s correct for just one, which is
all we’re trying to solve here.

https://codereview.appspot.com/7370046/diff/2001/input.py
File input.py (right):

https://codereview.appspot.com/7370046/diff/2001/input.py#newcode2617
input.py:2617: RemoveSelfDependencies(targets)
I think this should happen before any dependency_sections processing.

[glider1, 2013-02-27 13:15:46]

I've added a positive and a negative test for self dependencies, but the
negative one (real-self-dependency) fails with the following error report:

gyp: Some targets not reachable, cycle in dependency graph detected:
real-self-dependency/dep.gyp:dep#target
real-self-dependency/self_dependency.gyp:a#target

Are there death tests in gyp? I haven't managed to find any support for them.

https://codereview.appspot.com/7370046/diff/2001/input.py
File input.py (right):

https://codereview.appspot.com/7370046/diff/2001/input.py#newcode2617
input.py:2617: RemoveSelfDependencies(targets)
On 2013/02/25 18:49:43, Mark Mentovai wrote:
> I think this should happen before any dependency_sections processing.

Done.

[Mark Mentovai, 2013-02-27 13:50:37]

I don’t know if there are death tests in here, but the code LGTM.

[glider1, 2013-02-27 13:57:46]

On 2013/02/27 13:50:37, Mark Mentovai wrote:
> I don’t know if there are death tests in here, but the code LGTM.

I'll delete real-self-dependency then and move the test to self-dependency/

[glider1, 2013-02-27 14:10:33]

PTAL.
If you're fine with the code, can you please land it?

[Mark Mentovai, 2013-02-27 14:15:37]

Committed r1589