Bring some OOP and sanity to gclient.py.

- Bring some OOP and sanity to gclient.py.

- Changed the algorithm from breadth-first to depth-first.
- Added infinite recursion support.
- Fixed From() of From() dependency.
- Fixed cross solution custom deps aliasing.
- Removed support code for old .gclient_entries format.
- Added revinfo --actual, --snapshot is unchanged.
- Removed IsGitCheckout() in favor for a direct check for .git presence.

TEST=all test pass

(Superseeded by http://codereview.chromium.org/2867047)

[M-A Ruel, 2010-06-07 02:19:23]

... and removed more code than I added along the way.

The smoke tests all pass, I limited recursion to 2 levels since that is what
gclient was doing. It still doesn't resolve many gclient git issues.

[M-A Ruel, 2010-06-09 01:29:03]

ping, Nasser is on vacation.

[Mandeep Singh Baines, 2010-06-09 14:44:37]

This is a huge codereview which addresses multiple issues. I prefer one
codereview per independent change. Otherwise, I struggle to fit it all in my
head.

I've pointed out a couple of nits but I'm still getting my head around the
gclient.py changes.

http://codereview.chromium.org/2627007/diff/1/2
File gclient.py (right):

http://codereview.chromium.org/2627007/diff/1/2#newcode59
gclient.py:59: from types import NoneType
There is only only 1 use. Why not just types.NoneType?

http://codereview.chromium.org/2627007/diff/1/3
File gclient_scm.py (right):

http://codereview.chromium.org/2627007/diff/1/3#newcode680
gclient_scm.py:680: raise gclient_utils.Error('Directory %s is not present.' %
path)
Maybe this should be centralized to RunCommand?

http://codereview.chromium.org/2627007/diff/1/4
File gclient_utils.py (right):

http://codereview.chromium.org/2627007/diff/1/4#newcode296
gclient_utils.py:296: def FindGclientRoot(from_dir, filename='.gclient'):
Would anyone really every pass anything other than gclient?

If the intention is to collect all the '.gclient' strings in one place maybe use
a global constant.

http://codereview.chromium.org/2627007/diff/1/5
File tests/gclient_smoketest.py (right):

http://codereview.chromium.org/2627007/diff/1/5#newcode122
tests/gclient_smoketest.py:122: self.assertEquals(1223, len(result[0]))
It would be nice to see these magic numbers go away but I realize that is work
in progress.

[M-A Ruel, 2010-06-28 20:05:14]

Can you take a new look? The diff isn't as massive as before and I fixed all
outstanding issues (git-svn, pm incorrectly updating).

I have a hard time figuring out how to split this change further without writing
a fair amount of throw away code.

[M-A Ruel, 2010-06-30 19:58:41]

On 2010/06/28 20:05:14, Marc-Antoine Ruel wrote:
> Can you take a new look? The diff isn't as massive as before and I fixed all
> outstanding issues (git-svn, pm incorrectly updating).
> 
> I have a hard time figuring out how to split this change further without
writing
> a fair amount of throw away code.

Ping

[Mandeep Singh Baines, 2010-06-30 21:14:52]

There is a lot going on here. I really prefer changes which do one thing.

http://codereview.chromium.org/2627007/diff/21001/22001
File gclient.py (right):

http://codereview.chromium.org/2627007/diff/21001/22001#newcode205
gclient.py:205: if path[0] != "/":
path.beginswith("/")

http://codereview.chromium.org/2627007/diff/21001/22001#newcode460
gclient.py:460: %(solution_deps)s    },
Why?

http://codereview.chromium.org/2627007/diff/21001/22001#newcode468
gclient.py:468: %(solution_list)s]
Why?

http://codereview.chromium.org/2627007/diff/21001/22001#newcode551
gclient.py:551: text = 'entries = {\n%s}\n' % ''.join(result)
This change has nothing to do with the rest of the changes. It is easier to
review small independent changes versus a grab-bag of fixes.