Add git cl try that tries from Rietveld.

Add git cl try that tries from Rietveld.

Thin wrapper to process --bot and --testfilter to keep compability with gcl try
and git-try and pass the data directly to Rietveld.

R=rogerta@chromium.org
BUG=


Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=155176

[M-A Ruel, 2012-09-05 15:44:33]

Interestingly, trying a build on an invalid builder works;
http://codereview.chromium.org/8897024

Note that it's a Rietveld problem, not with this code.

Note that this makes revision default to HEAD instead of LKGR.  This is because
the poller defaults to HEAD. I can make a follow up CL to fix that client side.

[M-A Ruel, 2012-09-05 15:53:42]

On 2012/09/05 15:44:33, Marc-Antoine Ruel wrote:
> Note that this makes revision default to HEAD instead of LKGR.  This is
because
> the poller defaults to HEAD. I can make a follow up CL to fix that client
side.

Err, the client code shouldn't do that, since it must be the same for try jobs
triggered from the Web UI.

[Roger Tawa OOO till Jul 10th, 2012-09-06 01:33:59]

a few questions below.  I made similar comments on another CL, but then realized
that one was probably obsoleted by this one.

http://codereview.chromium.org/10916118/diff/1/git_cl.py
File git_cl.py (right):

http://codereview.chromium.org/10916118/diff/1/git_cl.py#newcode1576
git_cl.py:1576: builders_and_tests.setdefault(builder, []).append(tests)
should you split tests by ',' and then extend the list instead of just
appending?

http://codereview.chromium.org/10916118/diff/1/git_cl.py#newcode1579
git_cl.py:1579: builders_and_tests.setdefault(b, []).append('defaulttests')
the help text at line 1518 does not permit multiple bots, so either this elif is
not right or the help text is not right.  from a parsing point of view, seems
like sticking with the help text might be easiest?

http://codereview.chromium.org/10916118/diff/1/git_cl.py#newcode1592
git_cl.py:1592: patchset = properties['patchsets'][-1]
i think it should fail if there is no patchset.  before using 'git cl try', i
think the user should have already done a 'git cl upload', and the checkout
should be clean.  yes/no?

[M-A Ruel, 2012-09-06 07:35:12]

http://codereview.chromium.org/10916118/diff/1/git_cl.py
File git_cl.py (right):

http://codereview.chromium.org/10916118/diff/1/git_cl.py#newcode1576
git_cl.py:1576: builders_and_tests.setdefault(builder, []).append(tests)
On 2012/09/06 01:33:59, Roger Tawa wrote:
> should you split tests by ',' and then extend the list instead of just
> appending?

done

http://codereview.chromium.org/10916118/diff/1/git_cl.py#newcode1579
git_cl.py:1579: builders_and_tests.setdefault(b, []).append('defaulttests')
On 2012/09/06 01:33:59, Roger Tawa wrote:
> the help text at line 1518 does not permit multiple bots, so either this elif
is
> not right or the help text is not right.  from a parsing point of view, seems
> like sticking with the help text might be easiest?

done

http://codereview.chromium.org/10916118/diff/1/git_cl.py#newcode1592
git_cl.py:1592: patchset = properties['patchsets'][-1]
On 2012/09/06 01:33:59, Roger Tawa wrote:
> i think it should fail if there is no patchset.  before using 'git cl try', i
> think the user should have already done a 'git cl upload', and the checkout
> should be clean.  yes/no?

It's because of git cl patch.

[Roger Tawa OOO till Jul 10th, 2012-09-06 12:21:47]

Lgtm
On Sep 6, 2012 3:35 AM, <maruel@chromium.org> wrote:

>
>
http://codereview.chromium.**org/10916118/diff/1/git_cl.py<http://codereview....
> File git_cl.py (right):
>
>
http://codereview.chromium.**org/10916118/diff/1/git_cl.py#**newcode1576<http...
> git_cl.py:1576: builders_and_tests.setdefault(**builder, []).append(tests)
> On 2012/09/06 01:33:59, Roger Tawa wrote:
>
>> should you split tests by ',' and then extend the list instead of just
>> appending?
>>
>
> done
>
>
http://codereview.chromium.**org/10916118/diff/1/git_cl.py#**newcode1579<http...
> git_cl.py:1579: builders_and_tests.setdefault(**b,
> []).append('defaulttests')
> On 2012/09/06 01:33:59, Roger Tawa wrote:
>
>> the help text at line 1518 does not permit multiple bots, so either
>>
> this elif is
>
>> not right or the help text is not right.  from a parsing point of
>>
> view, seems
>
>> like sticking with the help text might be easiest?
>>
>
> done
>
>
http://codereview.chromium.**org/10916118/diff/1/git_cl.py#**newcode1592<http...
> git_cl.py:1592: patchset = properties['patchsets'][-1]
> On 2012/09/06 01:33:59, Roger Tawa wrote:
>
>> i think it should fail if there is no patchset.  before using 'git cl
>>
> try', i
>
>> think the user should have already done a 'git cl upload', and the
>>
> checkout
>
>> should be clean.  yes/no?
>>
>
> It's because of git cl patch.
>
>
http://codereview.chromium.**org/10916118/<http://codereview.chromium.org/109...
>

[commit-bot: I haz the power, 2012-09-06 12:35:58]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/maruel@chromium.org/10916118/6001

[commit-bot: I haz the power, 2012-09-06 12:38:30]

Change committed as 155176