rebaseline_server: deprecate --expectations-dir option in advance of git transition

rebaseline_server: deprecate --expectations-dir option in advance of git transition
(SkipBuildbotRuns)

R=bsalomon@google.com

Committed: https://code.google.com/p/skia/source/detail?r=12351

[epoger, 2013-11-21 17:08:37]

Ready for review at patchset 1.  (Nominating Brian as reviewer, but if someone
else wants to take it feel free)

Please let me know if:
1. you are aware of any uses of --expectations-dir we need to accommodate
2. my understanding of git limitations is correct

https://codereview.chromium.org/81243002/diff/1/gm/rebaseline_server/server.py
File gm/rebaseline_server/server.py (right):

https://codereview.chromium.org/81243002/diff/1/gm/rebaseline_server/server.p...
gm/rebaseline_server/server.py:135: self._expectations_repo =
_create_svn_checkout(
see comments below

https://codereview.chromium.org/81243002/diff/1/gm/rebaseline_server/server.p...
gm/rebaseline_server/server.py:456: 'Directory under which to find GM
expectations; '
Alternatively, we could allow the user to specify a path to git checkout root
containing expectations (a directory within which expectations can be found at
"expectations/gm").  By default, the path to git checkout root would be "../..",
and thus we would use the expectations within the same checkout as this script.

Two problems with that:

1. In order for _create_svn_checkout() at line 135 to check out expectations
into that other directory, it would need to download the entire history of the
entire Skia git repo.

2. The meaning of the option would have to change.  (The caller would have to
specify the path to git checkout root, not "expectations/gm" within the
checkout.)

[bsalomon, 2013-11-21 17:31:21]

deprecating, lgtm. Finer points of expectations in git are utm (unknown to me).

[epoger, 2013-11-21 17:33:35]

Leon- can you please weigh in on whether my understanding of git limitations is
correct here?

[scroggo, 2013-11-21 18:05:11]

On 2013/11/21 17:33:35, epoger wrote:
> Leon- can you please weigh in on whether my understanding of git limitations
is
> correct here?

You are correct.

I think it's fine to deprecate, but I don't understand the use case of setting a
different expectations directory.

[epoger, 2013-11-21 18:11:17]

On 2013/11/21 18:05:11, scroggo wrote:
> You are correct.

Thanks for checking.

> 
> I think it's fine to deprecate, but I don't understand the use case of setting
a
> different expectations directory.

One use case could be: you want to run rebaseline_server using --reload so that
it periodically updates GM expectations/actuals, but in such a way that it does
not pollute your local checkout.  (Keep those expectations in a temporary
directory, say)

Happily, this seems rare enough that we can remove the option.

[epoger, 2013-11-21 18:11:39]

Committed patchset #1 manually as r12351 (presubmit successful).