CBuildbot - Adds ability to only packages given by the sourcestamp.

CBuildbot - Adds ability to only rev packages given by the sourcestamp.

The buildbot drops a source stamp (repo1.git@commit1 repo2.git@commit2 ...).
If CBuildbot finds the source stamp, it uses it and revs only those packages.
If this file is either missing or set to "None", it assumes a Force Build
has been pressed and keeps the old behavior (attempts to mark all).

TEST=Ran with 3 states of revisions file.
BUG=5006

[sosa, 2010-08-23 22:55:17]

Also to Scottz

[sosa, 2010-08-24 16:56:54]

Ping.

On Mon, Aug 23, 2010 at 3:55 PM,  <sosa@chromium.org> wrote:
> Also to Scottz
>
> http://codereview.chromium.org/3163030/show
>

[Mandeep Singh Baines, 2010-08-24 22:17:21]

There is a lot going on here. Any chance this could be broken up?

[Mandeep Singh Baines, 2010-08-24 23:11:13]

http://codereview.chromium.org/3163030/diff/3001/4001
File bin/cbuildbot.py (right):

http://codereview.chromium.org/3163030/diff/3001/4001#newcode106
bin/cbuildbot.py:106: repo_cmd = "repo manifest -o /dev/stdout".split()
Why not "repo manifest -o -"?

http://codereview.chromium.org/3163030/diff/3001/4001#newcode200
bin/cbuildbot.py:200: for revision in revision_list:
for package, commit in revision_list:

http://codereview.chromium.org/3163030/diff/3001/4001#newcode267
bin/cbuildbot.py:267: revisions = 'None'
Why no just a regular None.

http://codereview.chromium.org/3163030/diff/3001/4001#newcode280
bin/cbuildbot.py:280: rev_all = (revisions == 'None')
rev_all seems unnecessary. You could simply test if revisions == None

[sosa, 2010-08-24 23:26:45]

PTAL.

Addressed all comments and added a unit test for _ParseRevisionString.  Have
tested _CreateDictionary manually.  Looking into adding mox for unittests for it
later.

http://codereview.chromium.org/3163030/diff/3001/4001
File bin/cbuildbot.py (right):

http://codereview.chromium.org/3163030/diff/3001/4001#newcode106
bin/cbuildbot.py:106: repo_cmd = "repo manifest -o /dev/stdout".split()
On 2010/08/24 23:11:13, Mandeep Singh Baines wrote:
> Why not "repo manifest -o -"?

Didn't know about '-o -'.  The documentation mentions it but doesn't describe
the behavior.  Done.

http://codereview.chromium.org/3163030/diff/3001/4001#newcode200
bin/cbuildbot.py:200: for revision in revision_list:
On 2010/08/24 23:11:13, Mandeep Singh Baines wrote:
> for package, commit in revision_list:

Done.

http://codereview.chromium.org/3163030/diff/3001/4001#newcode267
bin/cbuildbot.py:267: revisions = 'None'
On 2010/08/24 23:11:13, Mandeep Singh Baines wrote:
> Why no just a regular None.
Buildbot copies a 'None' into a file under Force Build.  Better to emulate the
same behavior o/w I have to check -- if revisions and revisions != 'None'

http://codereview.chromium.org/3163030/diff/3001/4001#newcode280
bin/cbuildbot.py:280: rev_all = (revisions == 'None')
I had this before but changed it so it's clear that rev_all == revisions ==
'None' since rev_all is much more explanatory.

On 2010/08/24 23:11:13, Mandeep Singh Baines wrote:
> rev_all seems unnecessary. You could simply test if revisions == None

[Mandeep Singh Baines, 2010-08-24 23:49:08]

LGTM. One minor nit. Your call if you want to address or not.

http://codereview.chromium.org/3163030/diff/3001/4001
File bin/cbuildbot.py (right):

http://codereview.chromium.org/3163030/diff/3001/4001#newcode267
bin/cbuildbot.py:267: revisions = 'None'
On 2010/08/24 23:26:46, sosa wrote:
> On 2010/08/24 23:11:13, Mandeep Singh Baines wrote:
> > Why no just a regular None.
> Buildbot copies a 'None' into a file under Force Build.  Better to emulate the
> same behavior o/w I have to check -- if revisions and revisions != 'None'
> 

OK

http://codereview.chromium.org/3163030/diff/3001/4001#newcode280
bin/cbuildbot.py:280: rev_all = (revisions == 'None')
On 2010/08/24 23:26:46, sosa wrote:
> I had this before but changed it so it's clear that rev_all == revisions ==
> 'None' since rev_all is much more explanatory.
> 
> On 2010/08/24 23:11:13, Mandeep Singh Baines wrote:
> > rev_all seems unnecessary. You could simply test if revisions == None
> 
> 

Its still clear withou rev_all because the else case is _UprevAllPackages()

[sosa, 2010-08-25 01:08:43]

Done.  Pushing