Add ability to rebaseline skimage to rebaseline.py

Add ability to rebaseline skimage to rebaseline.py

Skip pattern matching for skimage, since the skimage results do not
match the pattern.

Automatically fill in expectations_root and actuals_base_url when
--skimage is specified.

Requires a change to move the expectations files to the right place.
Depends on https://codereview.chromium.org/26734006

(SkipBuildBotRuns)

BUG=skia:1466
R=epoger@google.com

Committed: https://code.google.com/p/skia/source/detail?r=11717

[epoger, 2013-10-10 15:05:08]

https://codereview.chromium.org/26666004/diff/4001/tools/rebaseline.py
File tools/rebaseline.py (right):

https://codereview.chromium.org/26666004/diff/4001/tools/rebaseline.py#newcod...
tools/rebaseline.py:180: from_trybot=False, skip_pattern_matching=False):
please update the docstring above

https://codereview.chromium.org/26666004/diff/4001/tools/rebaseline.py#newcod...
tools/rebaseline.py:221: elif filepath.startswith('gs:'):
As discussed live: I think it's better to download files from Google Storage
using HTTP instead of gsutil.  (Assuming that the files have been uploaded as
publicly readable)

1. no need to install gsutil package
2. no need for gsutil keys

https://codereview.chromium.org/26666004/diff/4001/tools/rebaseline.py#newcod...
tools/rebaseline.py:443: args.actuals_base_url =
'gs://chromium-skia-gm/skimage/actuals'
I think it will be surprising that these args are ignored, depending on the
settings of other args.

Maybe you could do one of these instead?
- fail if the user specifies conflicting args
OR
- if args.skimage has been set then use these as DEFAULTS instead of ignoring
any user-specified values

... and then note the behavior in the description of the affected args

[scroggo, 2013-10-10 19:32:35]

https://codereview.chromium.org/26666004/diff/4001/tools/rebaseline.py
File tools/rebaseline.py (right):

https://codereview.chromium.org/26666004/diff/4001/tools/rebaseline.py#newcod...
tools/rebaseline.py:180: from_trybot=False, skip_pattern_matching=False):
On 2013/10/10 15:05:09, epoger wrote:
> please update the docstring above

Done.

https://codereview.chromium.org/26666004/diff/4001/tools/rebaseline.py#newcod...
tools/rebaseline.py:221: elif filepath.startswith('gs:'):
On 2013/10/10 15:05:09, epoger wrote:
> As discussed live: I think it's better to download files from Google Storage
> using HTTP instead of gsutil.  (Assuming that the files have been uploaded as
> publicly readable)
> 
> 1. no need to install gsutil package
> 2. no need for gsutil keys

Done.

https://codereview.chromium.org/26666004/diff/4001/tools/rebaseline.py#newcod...
tools/rebaseline.py:443: args.actuals_base_url =
'gs://chromium-skia-gm/skimage/actuals'
On 2013/10/10 15:05:09, epoger wrote:
> I think it will be surprising that these args are ignored, depending on the
> settings of other args.
> 
> Maybe you could do one of these instead?
> - fail if the user specifies conflicting args

Done, for --tests and --configs.

> OR
> - if args.skimage has been set then use these as DEFAULTS instead of ignoring
> any user-specified values

For these, I have now set them to the skimage values ONLY if they are not
already set to the defaults.

One result of the current patch is that if the user specified:

./tools/rebaseline.py --skimage --actuals-base-url '<the real default value>' it
would get overwritten... (I wasn't sure how I could differentiate between 'user
left it unset, so it is set to the default' and 'user specified the default')
That said, this use case is illogical (rebaseline skimage using gm actual
results).

> 
> ... and then note the behavior in the description of the affected args

Done.

[epoger, 2013-10-10 19:51:39]

LGTM at patchset 3, with or without these changes

https://codereview.chromium.org/26666004/diff/10001/tools/rebaseline.py
File tools/rebaseline.py (right):

https://codereview.chromium.org/26666004/diff/10001/tools/rebaseline.py#newco...
tools/rebaseline.py:184: from_trybot=False, skip_pattern_matching=False):
Please add from_trybot to the docstring (I know you didn't add the param, but
while you're in there...)

https://codereview.chromium.org/26666004/diff/10001/tools/rebaseline.py#newco...
tools/rebaseline.py:200: if skip_pattern_matching:
Maybe, instead of adding the new parameter that can conflict with the two
existing ones...

if self._tests or self._configs:
  self._image_filename_re = re.compile(gm_json.IMAGE_FILENAME_PATTERN)
else:
  self._image_filename_re = None

If you do this, you'll have to make sure _tests and _configs are None when you
construct the Rebaseliner object.  (You could do that within your "if
args.skimage" block)

[scroggo, 2013-10-10 20:18:58]

https://codereview.chromium.org/26666004/diff/10001/tools/rebaseline.py
File tools/rebaseline.py (right):

https://codereview.chromium.org/26666004/diff/10001/tools/rebaseline.py#newco...
tools/rebaseline.py:184: from_trybot=False, skip_pattern_matching=False):
On 2013/10/10 19:51:39, epoger wrote:
> Please add from_trybot to the docstring (I know you didn't add the param, but
> while you're in there...)

Done.

https://codereview.chromium.org/26666004/diff/10001/tools/rebaseline.py#newco...
tools/rebaseline.py:200: if skip_pattern_matching:
On 2013/10/10 19:51:39, epoger wrote:
> Maybe, instead of adding the new parameter that can conflict with the two
> existing ones...
> 
> if self._tests or self._configs:
>   self._image_filename_re = re.compile(gm_json.IMAGE_FILENAME_PATTERN)
> else:
>   self._image_filename_re = None
> 
> If you do this, you'll have to make sure _tests and _configs are None when you
> construct the Rebaseliner object.  (You could do that within your "if
> args.skimage" block)

Done. I like this much better!

[epoger, 2013-10-10 20:36:21]

lgtm

[scroggo, 2013-10-10 21:09:28]

Committed patchset #6 manually as r11717 (presubmit successful).