skimage: if --readPath points at an empty dir, create an empty expectations file

skimage: if --readPath points at an empty dir, create an empty expectations file

TBR=scroggo

Committed: https://code.google.com/p/skia/source/detail?r=9642

After being reverted, this has been replaced by the similar https://codereview.chromium.org/17101029/

[epoger, 2013-06-17 19:34:47]

Apologies for the TBR commit, Leon, but this should fix the partial buildbot
breakage after https://code.google.com/p/skia/source/detail?r=9641 .

We can come up with a different solution if you prefer...

https://codereview.chromium.org/17294006/diff/1/tools/tests/skimage/output-ex...
File tools/tests/skimage/output-expected/empty-dir/expectations.json (right):

https://codereview.chromium.org/17294006/diff/1/tools/tests/skimage/output-ex...
tools/tests/skimage/output-expected/empty-dir/expectations.json:2:
"actual-results" : {
This makes sense: there were no files in the input directory, so there are no
expectations.

https://codereview.chromium.org/17294006/diff/1/tools/tests/skimage/output-ex...
File tools/tests/skimage/output-expected/nonexistent-dir/expectations.json
(right):

https://codereview.chromium.org/17294006/diff/1/tools/tests/skimage/output-ex...
tools/tests/skimage/output-expected/nonexistent-dir/expectations.json:2:
"actual-results" : {
Currently, this case (--readPath pointing at a directory that does not exist)
yields identical results to being called for an empty directory.

We may want to change that-- it seems to me that readPath pointing at a
directory that doesn't exist is more of an error case.

[epoger, 2013-06-17 19:35:19]

Committed patchset #1 manually as r9642 (presubmit successful).

[epoger, 2013-06-18 02:56:28]

I have reverted this as https://code.google.com/p/skia/source/detail?r=9647 ,
since my fix to the buildbot scripts (
https://code.google.com/p/skia/source/detail?r=9645 ) makes it unnecessary.

Leon, if you think the change (or something similar) is a good idea, we can
certainly commit it again.  In the meanwhile, since you never had a chance to
review it, I have backed it out.

[scroggo, 2013-06-19 15:34:09]

https://codereview.chromium.org/17294006/diff/1/tools/skimage_main.cpp
File tools/skimage_main.cpp (right):

https://codereview.chromium.org/17294006/diff/1/tools/skimage_main.cpp#newcod...
tools/skimage_main.cpp:518: // FLAGS_readPath[i] is an empty or nonexistent
directory,
I am okay with the spirit of this change, but this else case is supposed to
cover someone calling

skimage --readPath image.png

With this change, that use case is no longer supported.

[epoger, 2013-06-19 15:54:15]

https://codereview.chromium.org/17294006/diff/1/tools/skimage_main.cpp
File tools/skimage_main.cpp (right):

https://codereview.chromium.org/17294006/diff/1/tools/skimage_main.cpp#newcod...
tools/skimage_main.cpp:518: // FLAGS_readPath[i] is an empty or nonexistent
directory,
On 2013/06/19 15:34:09, scroggo wrote:
> I am okay with the spirit of this change, but this else case is supposed to
> cover someone calling
> 
> skimage --readPath image.png
> 
> With this change, that use case is no longer supported.

Yup.  It is handy that --readPath can take either folders or files... but it
should probably handle empty directories differently than it does now (it acts
like the folder was a file it couldn't decode).

Anyway, happily, I don't need this change at present...